# Tool reference

> Every tool your own agent can call over MCP, grouped by the scope that opens it, with what each one does, how it answers, the recipes your agent can follow and the documentation it reads there as resources.

sourcesignal documentation, Spot and agents. Updated 2026-10-19. The page for people: https://www.sourcesignal.ai/docs/spot-and-agents/agent-tool-reference. Every article: https://www.sourcesignal.ai/llms.txt

This page lists every tool your own agent can call through the endpoint at `/api/mcp`: 215 tools under 5 scopes, 18 recipes it can follow, and this documentation, which it reads there as resources. It is built from the same list the endpoint serves. A tool that belongs to a feature your account does not have, such as the content plan, may be missing from your agent's list or say why it cannot run.

Your agent sees only the tools its token's scopes open. A tool outside them is not refused: it is missing from the list your agent receives. Its first call is `start`, which says who it acts for and which projects the token reaches. To make a token, see [Connect your own agent over MCP](https://www.sourcesignal.ai/docs/spot-and-agents/connect-your-own-agent-over-mcp.md). For limits, waits and long answers, see [How the door answers](https://www.sourcesignal.ai/docs/spot-and-agents/how-the-door-answers.md).

## How to read an entry

Each entry gives the name your agent calls, the title your agent sees, the first lines of what the tool tells your agent, and how it answers:

- **Reads and answers at once.** It reads what you can open in the app and changes nothing.
- **Answers with a link into the app.** It hands back the screen to open, never a change.
- **Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.** Nothing changes until you press Apply on the card, unless your setting under Settings, Agents applies that kind of change at once, or a mandate you approved names the tool, by name or by its kind of change: then it applies within the limits they set and, where the change can be taken back, you can take it back for a day. See [Change cards: apply and undo](https://www.sourcesignal.ai/docs/spot-and-agents/change-cards-apply-and-undo.md) and [Agent scopes and limits](https://www.sourcesignal.ai/docs/spot-and-agents/agent-scopes-and-limits.md).
- **Leaves a card for you to apply, or applies at once where your setting says so. No mandate reaches it.** It follows your setting like any other change, but no mandate lets your agent make it on its own: it acts on your account, such as renaming it, and a mandate covers one project, or it is an email to a publisher, which goes at once only where you set "Emails and posts to publishers" to Apply at once. Every email still passes the same checks as yours.
- **Always leaves a card for you to apply.** No setting and no mandate lets your agent apply it on its own: what you confirm happened outside the platform, such as a win or an email you sent yourself, a mandate itself, and the other changes that always wait for you.
- **Takes back your agent's own change, or withdraws its own card.** Your agent can put back a change it made on its own, for a day where the change can be taken back, or withdraw a card it left that still waits for you. It never touches a change you applied or a card you decided.
- **Uses the AI allowance.** It runs the platform's own AI for the account: `outreach_prompt_suggest`, `sequence_template_suggest`, `steer_compile`, `keywords_suggest`, `spot_plan`, `spot_ask`. No plan limits that allowance today: the work is counted for the account, not capped.

## Read your projects

Scope `project:read`, 55 tools. The projects you can open: the pages the engines cite, the deck, outreach, tracking and setup, with a brief of each project and the reasons behind a score or a card. Read only. Any member can hold it.

- `project_brief`: **Project brief.** The project in one read: identity, tracking, the funnel counts, the conversations waiting, outreach health, what is unfinished in setup, the plan and the rules in force, and Autopilot outlines. *Reads and answers at once.*
- `search`: **Search the project.** Find pages, tracked prompts, keywords and competitors in this project, the way the Ctrl K palette searches. *Reads and answers at once.*
- `open_view`: **Open a view.** Build an in-app link the answer renders as a button: a page of this project, the opportunities workspace filtered, one opportunity, one conversation, or a page with a dialog open. *Answers with a link into the app.*
- `glossary_lookup`: **Look up a term.** What a word means in this product, in the words you already see on screen: a metric, a status, a page type, a route, a deck group, the wins rule. *Reads and answers at once.*
- `compare_periods`: **Compare two periods.** The one way to say up, down or by how much. One metric over the last N complete days and the N days before them, with both values, the delta, the percentage and the direction, computed here. *Reads and answers at once.*
- `opportunities_find`: **Find opportunities.** Count and list the pages of this project that match a filter in the workspace's own vocabulary: status, page type, commercial class, content route, traffic light, prompt topic, engine, a score range, a quick filter, pinned or promoted, a text search. *Reads and answers at once.*
- `opportunity_explain`: **Explain one page.** What the page's own details drawer shows, for one opportunity: its status, score, page type, commercial class, route and traffic light in words; where the AI cites it; whether the page already names you; its size and its market; why engines cite it rather than you; the outreach route Autopilot takes for it and why; its contacts count, sequences and verified wins. *Reads and answers at once.*
- `deck_state`: **Today's deck.** The decision deck as it would be dealt right now: how many cards by group, the conversations behind it, and the first cards in dealt order with their page and group. *Reads and answers at once.*
- `pipeline_state`: **The pipeline.** What Autopilot is doing for this project: the feed and auto-promotion switches, the latest import, grading in flight, the research queue, the deck count, the send window and what is due, wins verified and pending, what the last 24 hours produced, replies in 24 hours, the scheduled jobs with their last run, the research pace, and the ten most recent jobs, each with its `job_id` and status. *Reads and answers at once.*
- `conversations_find`: **Find conversations.** The conversations of this project as the Inbox and the deck derive them from one set of facts: those waiting on your reply, due a reminder, taken over by a colleague, answered and waiting for a verdict, or live and waiting on the publisher. *Reads and answers at once.*
- `describe`: **Describe the data or a tool.** What can be read, and how a tool works. With no argument: the tables you can read, by area, with the rows you can see in each. *Reads and answers at once.*
- `query`: **Read rows.** One read over a table you can read: the columns to select, filters, up to three linked tables, an order, a limit and `next_cursor` for the next page; or an aggregate. *Reads and answers at once.*
- `read`: **Run a read by name.** Runs any read you can call by its name, with its arguments, and answers what it answers: an insight, a table read, a figure. *Reads and answers at once.*
- `metric`: **One named figure.** One named figure with the window it was counted over and what it means: wins, emails sent, replies, the reply rate, pages found, answers gathered, share of voice, mention rates, citations in five named units, data coverage and a page's scores. *Reads and answers at once.*
- `status`: **Status of a job, a result or a card.** Where something stands, by its id, waiting up to `wait_seconds` for it to settle when asked. *Reads and answers at once.*
- `jobs_find`: **Find jobs.** This project's jobs since a time, newest first: each with its id, its label, its status in words, who started it, when it was created and finished, and whether it failed with an error. *Reads and answers at once.*
- `next_steps`: **What to do first.** What to do first today on this project, ranked from what waits on you, each step with its count, the time on it where there is one, why it matters and the tool or page that acts on it. *Reads and answers at once.*
- `insights_since`: **What changed since.** What changed on this project since a cursor, for a reader coming back or an agent polling: the wins whose live day falls on or after it, the replies received, the placements found gone, the cards decided and how, the jobs that finished or failed, and the decisions taken by you, your teammates, your agents and Autopilot, apart. *Reads and answers at once.*
- `replies_find`: **Replies waiting.** Who is waiting on you and what they said: the latest reply of each conversation on this project, classified as the app read it, with the counts of every class. *Reads and answers at once.*
- `contacts_find`: **Find contacts.** The contacts you have on this project: the people found behind its pages, with their name, address and role as the app shows them, whether the address was checked and how it was found, when they were last contacted, and the page they belong to. *Reads and answers at once.*
- `placements_find`: **Placements won.** Every placement won on this project: the page, how it was won, its type, the day it went live, whether it is still live, the monitoring facts and the answers that cited it after it went live. *Reads and answers at once.*
- `outreach_yield`: **When replies come.** When replies come on this project's outreach, over the emails sent in the last N days: by step, by template and by the hour the answered email left, in the project's send window time zone. *Reads and answers at once.*
- `what_worked`: **What followed your pitches.** Which pitches were followed by movement, over the pages pitched in the last N days: how many were pitched, how many have been measured since, how many were followed by more mentions or citations on your tracked prompts, and how many went live afterward. *Reads and answers at once.*
- `mailbox_health`: **Mailbox health.** Sending health of the mailboxes this project sends through: each one's address and status in words, its daily limit and what it sent today in its own day, the emails sent and bounced over the last 30 days with the bounce rate, whether it is held from sending, and the pitches on it that are paused or need attention. *Reads and answers at once.*
- `spend_find`: **What was spent.** What Autopilot and your team's presses used on this project over the last N days: the presses by the action pressed and the metered jobs by type, Autopilot's apart from the ones a person queued. *Reads and answers at once.*
- `agent_activity`: **What your agents did.** What your agents did on this project over the last N hours: the changes an agent applied on its own under a mandate you approved, with the ones that can still be taken back and until when; the cards it left that wait for you, with when the first expires; and what became of the others. *Reads and answers at once.*
- `clients_health`: **Your clients at a glance.** Across every project you can see, the wins first: the placements won in the last 30 complete days, counted as the console counts them, in all and per project. *Reads and answers at once.*
- `prompts_find`: **Find prompts.** The tracked prompts of this project with the facts the Prompts page shows for the last 30 days: how many answers there were, how many named you, every competitor they named with how many answers, the one named most, and where you were placed. *Reads and answers at once.*
- `competitors_movement`: **Competitor movement.** Which competitors gain or lose share of voice on your tracked prompts, and where: each roster brand's share over the last N weeks against the N before, the change in points and in points a week, and the engine and the topic where it moved most, biggest movers first, with your own share beside them. *Reads and answers at once.*
- `visibility_breakdown`: **Share of voice breakdown.** Why your share of voice on your tracked prompts moved: the report's share over the last N complete days against the N before, the change split by engine and by topic, and the brands whose share rose or fell over the same days, in association words. *Reads and answers at once.*
- `tracking_health`: **Tracking health.** Whether a dip was an outage or a quiet feed, not fewer mentions: the Ahrefs feed's checks by day, the answers the report counts by day against the usual, the days with none, whether the feed is failing now in the banner's own words, whether the project is paused, and each engine's state in the settings card's words. *Reads and answers at once.*
- `engine_rates`: **Engine mention rates.** How often each AI engine names you, with a margin: per engine, the stored answers read in the window, how many name you, the rate and its 95% Wilson interval, in percent; and the same over every engine. *Reads and answers at once.*
- `score_history`: **Score history of a page.** Why one page's score reads what it does and how it moved lately: its priority and its score before ranking now, the parts of its current scoring and when it ran, the score before ranking of each earlier scoring with when it held, the change since the window began, and any score set by hand. *Reads and answers at once.*
- `fanout_find`: **What engines searched.** What the AI engines searched while answering your tracked prompts: the fan-out searches stored with each answer in the window, grouped by their words and counted, most frequent first. *Reads and answers at once.*
- `competitor_pages`: **Pages citing a competitor.** Pages the AI answers cite when they name a competitor and leave you out: from the answers stored in the window, the ones naming a brand on your roster and not you, and every page of yours to pitch that such an answer cites, with how many of those answers cite it and which competitors they named. *Reads and answers at once.*
- `market_find`: **Market benchmark.** The market across AI engines from the Ahrefs benchmark, per brand and engine over a window: share of voice, and mentions, citations and impressions, for every engine together and each engine apart. *Reads and answers at once.*
- `data_correlate`: **Cross-tab two columns.** A cross-tab of two columns of one table you can read: an aggregate grouped by both, each cell a count of rows or one metric, with filters and `include_hidden` as `query` takes them. *Reads and answers at once.*
- `deck_patterns`: **Patterns in your deck decisions.** The decisions you keep making by hand in the Autopilot deck over the last N days: the cards you skipped or parked, grouped by what a standing steering rule could stop for you. *Reads and answers at once.*
- `promotion_outcomes`: **What followed each promotion rule.** Each auto-promotion rule of this project over the last N days: the pages it promoted, how many of them were followed by a reply and by a verified win, the reply rate with its denominator, whether the rule is on, its criteria in words and the link to the rules page. *Reads and answers at once.*
- `research_yield`: **What contact research found.** What contact research found for this project over the last N days: the publishers searched, how many gave an address, how many gave people but no address, how many gave nobody, the rate with its denominator, the contacts found by research and how many had an email, and the pages given contacts already on file for their publisher, by route. *Reads and answers at once.*
- `drafts_find`: **Find pitch drafts.** The pitch drafts of this project as the Sequences page holds them, by stage: `needs_review`, ready, parked or all. *Reads and answers at once.*
- `posts_find`: **Find posts.** The posts of this project as the Posts view holds them, one stage at a time: the drafts to do, the posts done or the ones skipped. *Reads and answers at once.*
- `tracking_reports_find`: **Find tracking reports.** The tracking reports the account's own Ahrefs key can see, each by its id and the name the settings card lists it under, whether this project tracks with it and whether the platform made it. *Reads and answers at once.*
- `pages_review`: **Read the pages behind rows.** For one to ten pages, what the review pass shows beside each row: the page's title, its summary, a snippet of the page around where the brand is named, whether the page already names the brand, with or without a link, and when the page was read. *Reads and answers at once.*
- `keywords_find`: **Find keywords.** The keywords of this project as the Keywords page counts them: how many are tracked, how many are archived, the 500 per project ceiling and the room left, how many sit at each buying stage, and up to 20 keywords, each with its market, stage, monthly volume, cost per click, difficulty and the prompts linked to it. *Reads and answers at once.*
- `competitors_find`: **Find competitors.** The competitors of this project as the Competitors page lists them: the roster the report compares you with, how many are tracked and how many were removed, and the top brands the discovery queue suggests from the AI answers, with the automatic adds setting. *Reads and answers at once.*
- `content_gaps_find`: **Find content gaps.** The Content area's Gaps: the topics of this project's content plan with no page of yours that AI answers cite, highest priority first, as the Gaps step lists them. *Reads and answers at once.*
- `content_articles_find`: **Find articles.** The Content area's Articles: the pages planned for this project and where each stands, newest move first, with counts by stage. *Reads and answers at once.*
- `content_improvements_find`: **Find improvements.** The Content area's Improvements: for each topic where your site already has a page, the stored comparison of that page against the page AI answers cite instead, in the plan's order. *Reads and answers at once.*
- `content_checks_state`: **Site checks.** The Content area's Technical step: the newest check of your own site against the public AI crawlers and analyzers, as the page shows it. *Reads and answers at once.*
- `content_piece_get`: **Get a page.** One page of the content plan as the download gives it, for an agent that will put it on the site or hand it on: part "summary", "markdown", or "json", or "outline", or "section". *Reads and answers at once.*
- `report_headline`: **Report headline.** The Reports overview's own figures for this project, from the same loader the page uses, so every number matches the report: its summary cards, the report's own summary sentences, the competitor ranking by share of voice and the topics that moved most. *Reads and answers at once.*
- `start`: **Start here.** Without `project_id`: who your agent acts for, the projects this token reaches by id, the plan's room, how every change lands, what Spot does and never does, and recipes for common questions. *Reads and answers at once.*
- `catalog`: **Every tool you can call.** Every entry your agent can reach with this token, your role and the features switched on in this deployment: each with its name, kind, a one-line summary, what it costs and how it would apply for your agent now. *Reads and answers at once.*
- `spot_explain`: **Explain why.** Why something is as it is, with the evidence on record: a page, a deck card, a card this token left or a refusal. *Reads and answers at once.*

## Suggest rules and a focus

Scope `project:steer`, 5 tools. Suggest steering rules, a focus for the deck and the daily decision budget. Each one waits for you to apply it. Any member can hold it.

- `steer_compile`: **Read a steering sentence.** Turns one sentence about what the deck should deal or what Autopilot should stop working on into typed steering terms, chosen from the project's own lists: the routes, the pitch groups, the page types, the topics and up to 50 matching prompts. *Reads and answers at once. Uses the AI allowance.*
- `steer_set_rule`: **Set a steering rule.** Proposes a standing rule for the project: routes switched off and pitch groups left out of the deck. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `steer_set_focus`: **Set a focus for today's deck.** Proposes a focus for your own deck: only some pitch groups, quick decisions only, and a page predicate that reads exactly as the Opportunities filter reads it. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `steer_end`: **End a rule or clear the focus.** Proposes ending one of the project's steering rules or clearing your focus. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `deck_set_budget`: **Set the day's budget.** Proposes a daily decision budget for your deck on this project: how many new pitches the deck deals today, on your own local day; conversations and content pages never count toward it. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*

## Make changes in your projects

Scope `project:work`, 148 tools. Make changes in your projects, such as parking pages, researching contacts or setting deck rules. Each change waits as a card for you to apply, unless your setting in Settings, Agents, or a mandate you approved, applies that kind of change at once. An email or a post goes at once only where you set "Emails and posts to publishers" to Apply at once, never under a mandate, and through the same checks as yours. Never billing, members or access, and never a permanent delete. Any member can hold it.

- `pipeline_run`: **Run the pipeline.** Proposes one pipeline run for this project now, the same run the Run button starts: grading of new pages, contact research, drafts for pages a rule promotes, page checks and the rollups. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `contacts_research`: **Research contacts.** Proposes contact research for pages of this project, the same job the Research contacts button queues: it searches the web for the people behind each page and saves the finds. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `nudges_draft`: **Draft nudges.** Proposes one short follow-up email, as a draft, for each page whose follow-up date is due, the same draft the deck's "Draft a nudge" writes: it continues the earlier pitch to the same person from the same mailbox and lands in Outreach for your approval; nothing sends on its own. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `opportunities_park`: **Park pages in bulk.** Proposes parking pages of this project, the same bulk move the opportunities table's bulk bar makes: parked pages leave the deck and the default views and can be brought back. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `opportunities_bring_back`: **Bring pages back.** Proposes bringing parked pages of this project back into play, the same bulk move the opportunities table's bulk bar makes: each returns as New or Approved, so Autopilot and the deck see it again. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `opportunities_promote`: **Promote pages.** Proposes promoting pages of this project to outreach, the same press as the opportunities table's Promote to outreach: for each page a background job finds a contact and writes a pitch draft that lands in Outreach for your review; nothing sends on its own. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `job_retry`: **Run a job again.** Proposes queuing a failed or canceled job of this project again, the way the Engine room advises for a stalled run: a pipeline run is queued as a fresh manual run, a contact research job for the same pages through the same gates. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `mailbox_reconnect`: **Reconnect a mailbox.** The sending mailboxes of this account and which of them need reconnecting, with the link to Settings, Mailboxes, where you press Reconnect. *Answers with a link into the app.*
- `opportunity_update`: **Edit a page.** Proposes changing a page's own fields, the same save as the page's detail sheet and the table's cells: its notes, its follow-up date, its owner, a hand-set priority score or its outreach angle. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `opportunities_pin`: **Pin pages.** Proposes pinning pages of this project, or unpinning them, the same pin as the opportunities table's and the bulk bar's. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `opportunity_override_set`: **Override a page's reading.** Proposes setting or clearing one human override on one page, the same pickers as the page's detail sheet: its commercial class, its content route, its traffic light or its PR authority value. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `promotion_rule_propose`: **Propose a promotion rule.** Proposes a standing auto-promotion rule for this project, or a change to one, the same save as the rules editor: a name, the pages it promotes in the editor's own vocabulary and whether it is on. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `research_window_set`: **Set the research pace.** Proposes the project's research pace, the same setting the deck's daily budget slider mirrors to the server: how many decisions a day you make, from which contact research is kept ahead of review. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `deck_types_set`: **Switch a deck decision type.** Proposes switching one kind of decision in or out of the Autopilot deck for the whole project, the same switch as the deck's Decisions menu: outreach, competitors or content. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `reply_triage`: **Triage a reply.** Proposes recording what a reply means, the same triage as the deck's reply card and the inbox: interested, agreed or lost. *Always leaves a card for you to apply.*
- `prompts_add`: **Track prompts.** Proposes tracking new prompts in this project, the same add as the Prompts page's bulk add: each a question as a buyer would ask an AI engine, with an optional group and tags, at most 50. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `prompts_remove`: **Remove prompts.** Proposes removing tracked prompts from this project, the same Remove as the Prompts page's: each prompt moves to Removed with its answers kept, the pages its answers found stay in your list, and Restore brings it back. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `prompts_restore`: **Bring back prompts.** Proposes bringing removed prompts back into tracking, the same Restore as the Prompts page's: each comes back with its answers, its pages are linked again from the answers already stored, and on a project whose Ahrefs report the platform manages, the report gets it back. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `prompt_rename`: **Edit a prompt.** Proposes editing one tracked prompt, the same save as the Prompts page's Edit: its words, and optionally its group and tags. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `prompts_import_ahrefs`: **Import prompts from Ahrefs.** Proposes reading this project's Ahrefs report prompt list now, the same import as the Prompts page's button: new prompts join within the plan's tracked prompts, prompts the report dropped move to Removed, prompts that came back return, and new ones are classified at once where the AI allowance has room. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `keyword_prompts_link`: **Link prompts to a keyword.** Proposes linking tracked prompts to one keyword, the same link as the Keywords page's: the prompts show under that keyword there, so a keyword with search demand is tied to the questions that stand for it. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `keyword_prompts_unlink`: **Unlink prompts from a keyword.** Proposes taking prompts off one keyword, the same unlink as the Keywords page's: the prompts stay tracked, and the page stops showing them under that keyword. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `competitor_update`: **Edit a competitor.** Proposes editing one competitor on this project's roster, the same Edit as the Competitors page's: its name, its website domain, or both, for example to correct a domain the suggestions guessed. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `competitors_remove`: **Remove competitors.** Proposes taking brands off this project's competitor roster, the same Remove as the Competitors page's: each leaves the roster the report compares you with, its history stays so past reports still count it, and adding it again brings it back. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `competitor_suggestions_accept`: **Add suggested competitors.** Proposes adding brands the suggestions found in your AI answers to this project's competitor roster, the same Add as the suggestion queue's: each by its name from `competitors_find`, with its website domain when the reader knows it; otherwise Autopilot works the site out when the reader applies, using the AI allowance, and a brand whose site cannot be worked out stays suggested. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `competitor_suggestions_dismiss`: **Dismiss suggested competitors.** Proposes dismissing brands from this project's competitor suggestions, the same Dismiss as the suggestion queue's: the suggestions stop offering them, and the answers that name them are kept. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `competitors_autoadd_set`: **Add competitors automatically.** Proposes the competitor suggestions' automatic adds for this project, the same setting as the suggestion queue's: a number of prompts or null to turn it off. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `engines_set`: **Change the AI engines.** Proposes which AI engines this project tracks, in priority order, the same save as the project settings' engine grid: engine ids such as chatgpt, gemini, `google_ai_overviews`, `google_ai_mode`, claude, perplexity, copilot or grok, within the plan's number of engines. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `project_markets_set`: **Change the target markets.** Proposes this project's target markets in priority order, the same save as the project settings' market chooser: market codes such as us, gb, ie or ww for worldwide, with US state codes in priority order while the United States is on the list. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `answers_check_now`: **Check for new answers now.** Proposes checking this project's tracking feed for new AI answers now, the same check as the Data tab's Run import now and the setup page's Check now, instead of waiting for the daily one: it reads the feed, lands what is new, and Autopilot sorts it on a run it starts. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `analysis_start`: **Write a data scientist report.** Proposes a new data scientist report for this project, the same Run report as the Data Scientist tab's: it reads the project's data, writes a headline, findings, key figures and caveats with the steps behind them, and replaces the tab's report when it lands. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `draft_edit`: **Edit a pitch draft.** Proposes changing the words of pitch drafts, the same Save as the Sequences page: one to ten emails by `step_id`, each its subject, its body or its wait. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `draft_rewrite`: **Rewrite a draft with AI.** Proposes rewriting a pitch draft with the platform's model, the same Regenerate as the review card and the deck: the whole draft or one email of it. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `reply_draft_set`: **Leave a reply ready.** Proposes leaving a reply ready in a conversation, in your own words: it waits in the conversation, the Inbox row says Reply ready, and the composer opens with it when you opens the thread. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `reply_prepare`: **Draft a reply with AI.** Proposes drafting a reply in a conversation with the platform's model, the same press as the composer's Draft a reply with AI: an answer when their message is the latest move, or a nudge when yours is the latest after a reply of theirs. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `outreach_prompt_set`: **Set the outreach angle.** Proposes changing the project's outreach prompt, the standing angle and guardrails every pitch draft is written with: the outreach brief, the project description, the sender identity, approved claims, what never to include, the model email, the voice, follow-up instructions and anything else. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `reply_rules_set`: **Set reply rules.** Proposes replacing the project's reply rules: the conditions the reply reader checks every new reply against, beside its built-in read of negative replies. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `fee_stance_set`: **Set how replies answer fee requests.** Proposes setting how the platform's suggested replies answer a publisher who asks to be paid: decline paid placements, counter or accept up to a ceiling, with the ceiling's amount and currency and one line a reply may use. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `pitch_drafts_create`: **Draft a pitch.** Proposes pitch drafts for one to five people at a page, the same as the enroll dialog: by contact id, from the project's sender or `mailbox_id`, written by the platform's model, from a sequence template of this project, or the way an existing draft on the page was written. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `no_reply_redraft`: **Try again after no reply.** Proposes the next pitch after a finished one nobody answered, the same as the foot of that sequence: `new_angle` writes one new email to the same person, threaded under the pitch they never answered and from the same mailbox; `other_person` writes the pitch again for someone else at the page, the way the first was written, from the same mailbox while it can still send. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `recipient_page_draft`: **Draft this page instead.** Proposes a pitch draft to a draft's recipient for another page of this project they are a contact on, one the drafting rule held back for them, the same as the draft window's Draft this page instead: the platform's model writes it, the page moves to outreach ready with its own draft, and their other drafts in play on other pages are set aside under Parked, so they get one pitch. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `recipient_page_pick`: **Pitch this page.** Proposes choosing which page a draft's recipient is pitched for, the same as the draft window's Pitch this page: this draft is kept and their other drafts in play on the project's other pages are set aside under Parked, so they get one pitch. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `page_draft_write`: **Write the page's own draft.** Proposes writing pages' own copy-and-send drafts with the platform's model, kept on each page apart from the pitch drafts: mode opener writes the first email, as the page's Draft with AI; mode sequence writes the first email and its follow-ups, replacing what a page holds, for ten pages a card. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `opportunities_recommend`: **Recommend the outreach angle.** Proposes the outreach angle for one to fifty pages of this project, the same as the opportunities table's Recommend: the platform's model reads the project, its tracked prompts and competitors, and writes one outreach type, angle and approach per page type onto the pages of that type, moving a new page to reviewed. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `drafts_park`: **Park drafts.** Proposes setting pitch drafts aside, or bringing them back, the same as the Sequences page's Park and Bring back: parked true moves drafts out of Needs review, Ready to launch and the deck into Parked; parked false brings them back, each to the stage it left. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `sequence_template_save`: **Save a sequence template.** Proposes a sequence template for this project, or a change to one, the same Save as the Templates editor: its name, the page type it writes for, its mode, the emails or the brief, whether it writes with the outreach prompt, and whether it is the default for its page type or the project. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `email_template_save`: **Save an email template.** Proposes a single-email template for this project, or a change to one, the same Save as the Templates editor: its page type, name, subject, body and whether it is the default for its page type. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `email_templates_seed`: **Add the default templates.** Proposes the default email template of every page type for this project, written by the platform's model, the same as the Templates screen's seed: a page type with no default gets one; with force, the defaults the platform added before are written again. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `auto_promotion_set`: **Switch auto-promotion.** Proposes switching the project's auto-promotion on or off, the same switch as Outreach, Set up, Rules: on, the promotion rules that are on promote the pages they match on every pipeline run, and each page promoted gets a contact found and a draft written for your review; off, nothing is promoted on its own. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `draft_recipient_set`: **Change a draft's recipient.** Proposes switching a pitch draft to another contact on the same page, the same as the draft window's recipient picker: by contact id. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `drafts_sender_set`: **Change the sender of drafts.** Proposes changing which connected mailbox pitch drafts send from, the same as the draft window's sender picker: one to two hundred drafts by sequence id, or with `from_mailbox_id` every draft of this project stranded on a mailbox that can no longer send. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `sequence_timing_set`: **Set when drafts send.** Proposes the timing of one to two hundred pitches of this project, the same as the Sends popover and the bulk bar's timing: `start_after`, `send_rules` and, for drafts only, delays. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `outreach_timing_set`: **Set the days and time new drafts send.** Proposes the days and time every new pitch draft of this project starts with, the same as the Sends popover's "For every draft in this project": the project's default timing, written onto the drafts already there too unless `apply_to_drafts` is false. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `timing_presets_set`: **Save or remove a timing preset.** Proposes saving a send-timing preset, or removing one, the same as the Sends popover's presets: a preset is offered in one press when timing a draft, and saving or removing one changes no draft. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `outreach_hold`: **Hold outreach.** Proposes pausing live pitches, the same as the tracker's Pause: up to two hundred by sequence id, or every pitch of this project sending now. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `outreach_stop`: **Stop sequences.** Proposes stopping live pitches for good, the same as the tracker's Stop: one to fifty by sequence id. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `follow_up_dates_set`: **Set follow-up dates.** Proposes the follow-up reminder date of one to two hundred pages of this project, or clears it, the same as a page's follow-up date and the table's bulk Set follow-up: the day the page comes up again to follow up. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `reply_reminder_set`: **Snooze or close a reply reminder.** Proposes snoozing a conversation's reply reminder or closing it for their latest message, the same as the deck's follow-up card. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `reply_reminder_days_set`: **Set reply reminder days.** Proposes the project's reply reminder setting, the same as Settings, Reply reminders, for an admin or owner: how many working days a conversation may wait on the publisher before it is marked "Nudge?" and dealt on the deck as a follow-up card, or 0 to switch reminders off. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `outreach_signoff`: **Sign off drafts.** Proposes signing off pitch drafts as read, the same as Mark reviewed on the Sequences page: one to fifty drafts by sequence id. *Leaves a card for you to apply, or applies at once where your setting says so. No mandate reaches it.*
- `outreach_launch`: **Launch drafts.** Proposes launching signed-off pitch drafts, the same as Launch on the Sequences page: one to fifty by sequence id. *Leaves a card for you to apply, or applies at once where your setting says so. No mandate reaches it.*
- `outreach_send_now`: **Send the next email now.** Proposes sending the next email of a launched pitch now, the same as Send now in the tracker: one pitch by sequence id. *Leaves a card for you to apply, or applies at once where your setting says so. No mandate reaches it.*
- `outreach_resume`: **Resume a sequence.** Proposes resuming a paused or answered pitch, the same as Resume in the tracker: one pitch by sequence id. *Leaves a card for you to apply, or applies at once where your setting says so. No mandate reaches it.*
- `outreach_resume_anyway`: **Resume into a conversation.** Proposes resuming a paused or answered pitch into a conversation, the same as Send anyway on the tracker's Resume: one pitch by sequence id, when they wrote to you or you wrote to them by hand, or when the pitch is held because they are talking to you on another thread or got an email from you for another page this week. *Always leaves a card for you to apply.*
- `reply_send`: **Send a reply.** Proposes sending a reply in a conversation, the same as Send reply in the Inbox: the reply that waits there, or your own words in body. *Leaves a card for you to apply, or applies at once where your setting says so. No mandate reaches it.*
- `step_retry_unconfirmed`: **Send an unconfirmed email again.** Proposes sending again an email the mail provider may already have delivered, the same as Send again in the tracker: one email by sequence id and step id. *Always leaves a card for you to apply.*
- `step_delivered_record`: **Record an email as delivered.** Proposes recording as delivered an email the mail provider may already have sent, the same as It was sent in the tracker: one email by sequence id and step id. *Always leaves a card for you to apply.*
- `draft_sent_record`: **Record a page draft as sent.** Proposes recording that you sent a page's copy-and-send draft from their own mail, the same as Mark as sent on the page: the page by opportunity id, which email, and the page's contact it went to when they know it. *Always leaves a card for you to apply.*
- `outreach_attempt_record`: **Record a contact made elsewhere.** Proposes recording that you contacted someone outside the platform, the same as Mark contacted on the contact, or Contacted on LinkedIn after no reply: the contact by id, how, and a short note. *Always leaves a card for you to apply.*
- `conversation_mark_won`: **Mark a conversation won.** Proposes marking a conversation won, the same as Mark as win in the Inbox and the deck: the conversation by sequence id, and the day the placement went live. *Always leaves a card for you to apply.*
- `post_mark_posted`: **Mark a post as posted.** Proposes recording that you posted a draft from their own account, the same as Mark as posted in the Posts view: the post by id, with the public link where it landed. *Always leaves a card for you to apply.*
- `suppression_lift`: **Lift a do-not-contact entry.** Proposes lifting one entry of your account's do-not-contact list, the same as Revoke under Settings, Do not contact: the entry by id. *Always leaves a card for you to apply.*
- `mailbox_send_settings_set`: **Set a mailbox's sending.** Proposes changing how one of the account's connected mailboxes sends, the same as its settings under Settings, Mailboxes: the mailbox by id, and any of the hours it sends in, the emails a day it sends at most, the least gap between two of its emails and the sign-off name new drafts from it use. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `contact_add`: **Add a contact.** Proposes adding a contact to a page, the same as the Recipients popover's add: a name or an address at least, with a role and notes if you have them. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `contact_update`: **Edit a contact.** Proposes editing a contact's name, address, role or notes, the same as the contact's own edit. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `addresses_check`: **Check addresses.** Proposes checking one to fifty contacts' addresses with Hunter, the same as Verify on a contact: each verdict is saved on the contact, and a pitch to an invalid address is blocked. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `contacts_suppress`: **Add to do-not-contact.** Proposes adding a contact's address to your account's do-not-contact list, the same as the contact's own button: scope email for that address, or domain for everyone at their publication. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `unmatched_mail_resolve`: **Attach or dismiss a suggested reply.** Proposes deciding an email the Inbox could only suggest for a conversation, the same as the Inbox's two buttons: attach makes it part of the conversation exactly as a matched reply, dismiss marks it not this conversation and it leaves the Inbox. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `conversation_contact_switch`: **Make the writer the contact.** Proposes making the person now writing in a conversation its contact, the same as the Inbox's button: they become a contact on the page from the address they wrote from, and the conversation points at them, so the list, the deck and your next reply follow. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `page_mention_recheck`: **Check the page again.** Proposes reading a page again, past its 30-day copy, for whether it already names or links your brand, the same as Re-check page on the page's sheet: the verdict is saved on the page, the one a draft reads too. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `post_draft`: **Draft a post.** Proposes drafting a post for a page with the platform's model, the same as Draft a post in the Posts view: a reply in a cited thread, a new post in its community, a directory listing's copy or a review site's profile copy. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `post_edit`: **Edit a post.** Proposes new words for a post draft under To do, the same as editing it in the Posts view: a reply's or a new post's text and title, or a listing's or a profile's copy by its own keys. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `posts_skip`: **Skip posts.** Proposes skipping one to fifty post drafts under To do, the same as Skip in the Posts view: each moves to Skipped and stays on the record, a pass on those words, never a verdict on the page. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `posts_reopen`: **Reopen posts.** Proposes moving one to fifty skipped or done posts back to To do, the same as Reopen in the Posts view: the words only. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `outreach_sender_set`: **Set who outreach goes out from.** Proposes who this project's outreach goes out from, the same Save as the project settings' outreach identity, for an admin or owner: the connected mailbox, the signature of that mailbox or its own default, "None" to send unsigned, and the number a publisher may use to invoice. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `outreach_hours_set`: **Set the hours outreach may go out.** Proposes the hours this project's outreach may go out, the same Save as the project settings' sending window, for an admin or owner: a timezone, the days and a first and last minute, or null to clear them so each mailbox's own hours apply alone. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `ahrefs_report_bind`: **Connect a tracking report.** Proposes connecting this project to a tracking report by its id, the same Save as the project settings' report id, or null to disconnect it. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `ahrefs_prompts_sync`: **Send your prompts to the tracking report now.** Proposes sending this project's tracked prompts to its tracking report now, the same Sync prompts now as the project settings' tracking card, instead of waiting for the next check: prompts added here are added to the report, prompts removed here come off it, and a prompt filed under another country is added under the project's. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `feed_reconnect`: **Reconnect the tracking feed.** Proposes repairing this project's feed of AI answers, the same as Retry on the feed banner, when answers stop coming in: a report the platform made has its engines and prompts brought back in line and is checked again, one gone from the Ahrefs account is made again with the prompts sent to it, and a report connected by hand is checked again as it is. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `brand_domain_set`: **Set the brand's website.** Proposes the brand's website for this project, a bare host such as acme.com. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `tracking_activate`: **Switch tracking on.** Proposes switching tracking on for this project, the same Switch tracking on as setup: a tracking report is made on Ahrefs for the brand's website, the questions it tracks are sent to it, and the feed's first read is queued, so the first answers land within minutes to a few hours. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `job_cancel`: **Stop a job.** Proposes stopping one of this project's jobs by its id from `jobs_find`, the same Cancel as the Engine room's. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `thread_decide`: **Decide a conversation card.** Proposes one of the deck's two answers on a conversation card: `nothing_to_add` or `keep_contact`. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `opportunities_regrade`: **Re-score pages.** Proposes reading pages again and re-scoring them under the project's own scoring version, the same Re-classify as the opportunities page's, 1 to 20 pages by id from `opportunities_find`. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `opportunities_group`: **Group and score pages with AI.** Proposes the opportunities page's bulk Group or Score on 1 to 50 pages by id from `opportunities_find`: step group sorts each page into its type with a one-line summary, and step score scores each page under the project's own scoring version. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `prompts_classify`: **Classify prompts again.** Proposes judging tracked prompts again for their buying stage and the brands they name, the same Classify as the Prompts page's, 1 to 50 prompts by id from `prompts_find`. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `competitor_suggestions_add_all`: **Add the top suggested competitors.** Proposes the suggestion queue's Add all: the first 50 companies the queue draws join this project's competitor roster, with their sites worked out by AI when the reader applies, and the category phrases in the queue join only by a site of their own cited in the answers. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `project_create`: **Create a project.** Proposes a new project for a client, the same create as the new-project form: its name, the brand as AI answers name it, the brand's website, the industry and the target markets in priority order, with US states while the United States is one. *Always leaves a card for you to apply.*
- `project_archive`: **Archive a project.** Proposes archiving this project, the same Archive as project settings': the project leaves everyone's project list and the plan's count of projects, and its data stays as it is. *Always leaves a card for you to apply.*
- `project_restore`: **Restore an archived project.** Proposes bringing this archived project back, the same Restore as the project list's: it rejoins everyone's project list and the plan's count of projects, and its tracked prompts rejoin the plan's pooled prompts. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `tracking_plan_propose`: **Propose questions and competitors to track.** Proposes a tracking plan for this project, the same plan the setup wizard drafts: the platform's AI reads the brand's site for its service areas, drafts the buyer questions AI engines are asked for each area, and proposes the competitors to compare against. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `prompts_propose`: **Propose more questions for a group.** Proposes more questions to track for one group of this project, 3 to 12 of them, 8 by default: the platform's AI drafts the questions buyers ask AI engines about it with the brand's context, and code leaves out the ones the project already tracks. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `org_rename`: **Rename the account.** Proposes a new name for the account, the same change as the name field on Settings, Organization: everyone on the account sees it in the switcher and on their settings. *Leaves a card for you to apply, or applies at once where your setting says so. No mandate reaches it.*
- `profile_update`: **Update your profile.** Proposes a change to your own profile, the same saves as the Profile tab: their first and last name, the company they work for, and their timezone. *Leaves a card for you to apply, or applies at once where your setting says so. No mandate reaches it.*
- `support_ask`: **Write to the support team.** Proposes a message to the platform's support team in your name, the same as writing in the support drawer: open a new conversation, answer in an existing one by its `thread_id`, or, with `feed_failure` true and a `project_id`, report that project's tracking feed failing with what the platform knows about the feed attached. *Always leaves a card for you to apply.*
- `support_resolve`: **Mark a support conversation solved.** Proposes marking one of your support conversations as solved, the same as Mark as solved in the support drawer: nobody is emailed, and writing in it again opens it again. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `support_rate`: **Rate a support answer.** Proposes your rating of the support team's answer in a solved conversation, the same as the rating in the support drawer: Great, Okay or Poor, with an optional note. *Always leaves a card for you to apply.*
- `keywords_add`: **Add keywords.** Proposes adding keywords to this project, the same add as the Keywords page's Add and bulk add: each term with an optional market and buying stage, and its monthly searches, cost per click and difficulty when known, at most 50. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `keywords_remove`: **Remove keywords.** Proposes removing keywords from this project, the same Remove as the Keywords page's: each keyword moves to the Archived tab with its prompt links kept, and Restore brings it back. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `keyword_update`: **Edit a keyword.** Proposes editing one tracked keyword by its id from `keywords_find`, the same Edit as the Keywords page's: its words, its market, its buying stage, and its monthly searches, cost per click and difficulty. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `keywords_restore`: **Restore keywords.** Proposes bringing archived keywords back to this project, the same Restore as the Keywords page's Archived tab: 1 to 50 keywords by id from `keywords_find`, each with its prompt links. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `keywords_import_csv`: **Import keywords from a CSV.** Proposes importing keywords from CSV text, the same Import CSV as the Keywords page's: a header row naming a keyword column, and optionally market, buying stage, monthly searches, cost per click and difficulty, at most 100,000 characters and 5,000 rows. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `keywords_suggest`: **Suggest keywords.** Proposes keywords found in the prompts this project tracks. One call to the platform's AI reads up to 10 tracked prompts and the searches AI engines ran while answering them; code then cleans each term, leaves out the ones that break the rules and marks the ones already tracked. *Always leaves a card for you to apply. Uses the AI allowance.*
- `competitors_add`: **Add competitors.** Proposes adding brands to this project's competitor roster, the same add as the setup page's and the tracking plan's: each with its name and website domain, at most 10. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_outline_start`: **Outline a page.** Proposes starting the outline of one page for a topic of the content plan, the same press as "Outline it" on a gap: a background job writes an outline from the answers that cite other sites, and it lands in Content, Articles for your review; nothing is published. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `site_checks_run`: **Check your site again.** Proposes checking your own site again against the public AI crawlers and analyzers, the same press as "Check again" on the Content area's Technical step: a background job reads the site's pages and the results replace the last run on that page. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `site_read_run`: **Read your site.** Proposes reading your own site for prompts and competitors to suggest, the same press as Rediscover on the Prompts and Competitors pages and "Read your site" in Content: a homepage read or a deep dive page by page. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_piece_mark_published`: **Mark a page as live.** Proposes recording where a page of the content plan went live, the same as "Published it?" on a downloaded page: the page's address on the site, or null to clear it. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_outline_again`: **Outline a page again.** Proposes outlining one page of the content plan again, the same press as "Outline again" on a ready outline and Retry on one that could not be finished: a background job writes a new outline from the page's frozen evidence, and it lands in Content, Articles for your review; the outline it replaces stays in the page's history, and nothing is published. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_outline_edit`: **Edit an outline.** Proposes saving your own outline over a page's ready outline, the same as "Edit outline" and Save: the whole outline in the shape `content_piece_get` part "outline" returns, with your changes. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_draft_start`: **Write the draft.** Proposes writing the draft of one page from its ready outline, the same press as "Write it": a background job writes the page section by section from the outline's frozen evidence, and the draft lands in Content, Articles for your review; nothing is published. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_draft_stop`: **Stop writing.** Proposes stopping a draft that is being written, the same press as "Stop writing": the page goes back to its ready outline, its job is canceled, and the sections written so far are kept in the page's history, not as the draft. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_section_rewrite`: **Rewrite a section.** Proposes writing one section of a page's draft again, the same press as Rewrite on a written section: a background job writes it again from the page's frozen evidence as the section's next version; the version it replaces stays in the section's history, and `content_section_edit` can put its words back. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_sections_retry`: **Retry the missing sections.** Proposes writing again only the sections of a page's draft that could not be written, the same press as Retry on a missing section: a background job writes those sections from the page's frozen evidence, and the rest of the draft stays as it is. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_section_edit`: **Edit a section.** Proposes saving your own words for one section of a page's draft, the same as the section editor's Save: the section's whole new Markdown and, if you change it, its heading. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_topic_decide`: **Set a topic aside, drop it or plan it again.** Proposes deciding one or more topics of the content plan, the same as the Gaps step's Not now, Drop this topic and Plan it again: `not_now` sets them aside until a day and they come back as gaps on their own then; drop takes them out of the plan, listed under Set aside with Restore; restore plans them again, voiding a set aside or a drop. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_piece_move`: **Move a page in the plan.** Proposes moving one page of the content plan, the same as Not now, Back, Discard and Restore in Content, Articles: `not_now` sets it aside until a day; back brings a page set aside back to where it stood; discard takes it out of the plan; restore brings a discarded page back. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_live_check`: **Check a live page again.** Proposes looking at a page of the content plan where it went live, now, the same as "Check again" on a live page: one fetch of the address, whether the client's own site answers it, and the AI answers for the project that cite it since it went live. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_plan_rank`: **Rank the content plan now.** Proposes ranking the project's content plan now, the same ranking the plan runs when a reader first opens it and after each data check: the topics read from the latest AI answers, matched against a light read of the client's own site, so Gaps and Improvements are current. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_autopilot_set`: **Set Autopilot outlines.** Proposes setting the project's Autopilot outlines, the same switch and weekly number as Content, Articles, Set up: on, from the next data check Autopilot outlines the highest priority gaps on its own, up to the weekly number, and each outline waits for your review; no draft is written without them. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_voice_read_run`: **Read your site's voice.** Proposes reading the client's own site for its voice, the same as "Read my site" in Content, Articles, Set up: up to six of its pages are read, the facts are counted by code, and the platform's model writes a short description. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_voice_accept`: **Accept the voice profile.** Proposes accepting the voice profile read from the client's site, the same as Approve on Set up's voice card, with your edits when you give them: the description of how they write, the terms the writer uses, the model page, whether em-dashes are kept and US or UK spelling. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `content_voice_forget`: **Forget the voice profile.** Proposes forgetting the voice profile read from the client's site, the same as Forget on Set up's voice card: outlines and drafts follow the voice in the outreach brief again. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `placement_record`: **Record a placement.** Proposes recording a live placement, the same save as Add manual win on the wins screen and Mark as won. *Always leaves a card for you to apply.*
- `placements_confirm_live`: **Confirm placements as live.** Proposes confirming agreed placements as live, the same pass as Confirm as live on the wins screen: each page is recorded as a live placement as you confirms it, with the day it went live, row by row, 1 to 200. *Always leaves a card for you to apply.*
- `placement_update`: **Edit a placement.** Proposes an edit of a placement, the same save as its Edit on the wins screen: the address, the title, the live date, the topic, the type, the team's note, the status or the linked prompts. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `placements_check`: **Check placements now.** Proposes checking placements now, the same press as Verify placements on the wins screen: each placement's page is read again for your brand, and whether it is live, at risk or gone follows what the check finds. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `share_link_create`: **Make a share link.** Proposes a public link, the same as Share report on the Overview or a briefing's Share. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `share_link_revoke`: **Turn off a share link.** Proposes turning a public link off, the same as Revoke in the Share report dialog or a briefing's Share: the link answers not-found from then on, and stays listed under Revoked. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `report_commentary_set`: **Write the report's commentary.** Proposes the agency commentary on the client report, the same save as Add agency commentary on the Overview: the words print on the Overview, on every share link and on the PDF, and the report's own numbers stay untouched. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `briefing_generate`: **Write a briefing.** Proposes writing a fresh briefing with the platform's model, the same Generate as on its panel: kind snapshot is the analyst's read of one snapshot, kind period the read of the last 7, 30 or 90 days against the window before, kind strategist the portfolio read of every page the project tracks. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `csv_import`: **Import a CSV.** Proposes importing a CSV as a new snapshot, the same upload as the Data tab's: kind opportunities, visibility or responses. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `traffic_sync_now`: **Pull traffic from Google now.** Proposes pulling this project's traffic from Google now, the same as Sync now on the Google traffic settings card. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `analysis_ask`: **Ask the data scientist.** Proposes asking the data scientist one question about this project's data, the same as Send in the analyst chat on the Data Scientist tab: the question joins the chat as your, and a background job reads the project's data and writes the answer there, with its findings and the steps behind them. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `mandate_propose`: **Ask for a mandate.** Asks you for a mandate on this project: what your agent may do on its own, what it will ask about first, decisions a day, the paid work a day, and how many days until it ends. *Always leaves a card for you to apply.*
- `deck_next`: **The next deck cards.** The decision deck as it is dealt for you right now: the next cards in dealt order, each with its key, kind, page, score, why it was dealt, its effort class and the decisions an agent may make on it, the steering it was dealt under, and the mandate in force with the decisions left today. *Reads and answers at once.*
- `deck_decide`: **Park or skip pages in the deck.** Decide one deck card by its key: park, skip, or undo. Inside a mandate you approved that names it, by name or by kind, and within its decisions a day, the decision is made at once and recorded as your agent's, under the mandate. *Leaves a card for you to apply, or applies at once where your setting or a mandate you approved says so.*
- `take_back`: **Take back your own change.** Takes back a change your agent made on its own, at once or under a mandate, within its day, or withdraws a card your agent left that still waits for you, by the `card_id` the change answered. *Takes back your agent's own change, or withdraws its own card.*

## Ask Spot

Scope `spot:ask`, 4 tools. Ask Spot for plans and answers about your projects. Each answer uses the account's AI allowance. Any member can hold it.

- `outreach_prompt_suggest`: **Suggest the outreach angle.** Suggests outreach prompt sections with the platform's model, the same as the editor's Suggest with AI: from the brand, its market, its topics and the pages it reached, live placements first, then agreed, then pitched. *Reads and answers at once. Uses the AI allowance.*
- `sequence_template_suggest`: **Suggest a sequence template.** Suggests a sequence template's emails with the platform's model, the same as the Templates editor's Draft with AI: for a page type, optionally from a brief of how to pitch and with the project's outreach prompt. *Reads and answers at once. Uses the AI allowance.*
- `spot_plan`: **Plan options for a goal.** For a goal and constraints: two or three options, each a plan from the catalog, a forecast from this project's own history as ranges with their sample size, what you must approve for it, and the trade-offs in words. *Answers within the call, or with a result your agent polls with `status`. Uses the AI allowance and counts toward your 30 questions an hour.*
- `spot_ask`: **Ask Spot.** An open question about this project, answered the way the panel answers you: a lead line, at most five facts each traced to a read, short plain sentences, the reads it used, and suggested next calls. *Answers within the call, or with a result your agent polls with `status`. Uses the AI allowance and counts toward your 30 questions an hour.*

## Read the account

Scope `org:read`, 3 tools. The account's projects, how each is doing, and what the account has used this month. For admins and owners. Read only. It works only on a token that reaches every project.

- `org_brief`: **Account brief.** The account in one read, for owners and admins: its projects, members by role, connected mailboxes by status, the plan and its phase, the features switched on in this deployment. *Reads and answers at once.*
- `org_projects_health`: **Projects' health.** One health line per active project of the account, for owners and admins: decisions waiting in its deck, the last pipeline run and whether it finished, jobs failed in the last 7 days, replies received in the last 7 days, wins verified in the last 30 days. *Reads and answers at once.*
- `org_usage`: **Plan usage.** The account's plan and its meters as Settings, Billing counts them, for owners and admins: each limit with what is used, the cap and when it resets, plus the plan name, its status and phase and the trial days left. *Reads and answers at once.*

## Recipes

A recipe is a short plan for a common question: the calls in order, each with what it returns. `start` hands them to your agent, and a client that lists prompts offers each one under its name. A recipe runs nothing by itself, and a change it leads to lands like any other change: as a card, or at once where your setting or a mandate says so. Your agent is offered a recipe only when its token opens every tool the recipe calls.

- `recipe_job_follow_up`: Did the work you started finish, and what did it produce.
- `recipe_invisible_prompts`: Questions where competitors are named and the brand is not.
- `recipe_unlinked_mentions`: Pages that name the brand without a link.
- `recipe_steer_the_deck`: Stop dealing a kind of page, see the rules, end one.
- `recipe_decision_history`: What was decided this week, take-backs included.
- `recipe_drafts_ready`: Which drafts are ready and what is left to fill.
- `recipe_log_placement`: Record a placement that was won.
- `recipe_first_today`: What to do first today.
- `recipe_since_last_look`: What changed since the last look.
- `recipe_share_drop`: Why share of voice dropped and what moved.
- `recipe_wins_ledger`: Every placement won, and what came after.
- `recipe_why_paused`: Why the project is paused or doing less.
- `recipe_outreach_on_its_own`: Run outreach on its own, short of sending.
- `recipe_new_client`: Set up a new client, from a website to a tracking plan.
- `recipe_write_article`: Write the page the content plan ranks first, from outline to draft.
- `recipe_client_report`: Prepare the monthly client report: the numbers, the commentary and a link.
- `recipe_placements_at_risk`: Find the placements at risk and check them now.
- `recipe_more_room`: More data or more relevant answers: what the plan has room for.

## Documentation

Your agent can read this documentation where it is connected, as resources, so a question about a screen or a status is answered from the same articles you read. A client that lists resources shows the index and every article; one that does not can read the same text at the addresses in [The docs for agents](https://www.sourcesignal.ai/docs/spot-and-agents/the-docs-for-agents.md).

- `sourcesignal://docs`: **Documentation index.** The product documentation for a model: one line per article with its summary and its markdown address, then the common questions and the main pages of the site. It says what a person sees on each screen, never how a page is scored.
- `sourcesignal://docs/{section}/{slug}`: **One documentation article.** One article as markdown: its title, summary, section, the day it was updated, the page for people, the body and the related articles. The index lists every section and slug.

## The change card

Every tool that prepares a change also names an interactive card. Agent apps that show interactive cards, such as Claude, ChatGPT and VS Code, draw it in the chat beside your agent's answer: the title, what it changes and the time left, with one action that opens the card here, where you apply or dismiss it.

- `ui://sourcesignal/change-card`: **Change card.** It loads nothing from the web and applies nothing. Apps that do not show interactive cards show your agent's answer and its link instead.

## Related

- [How the door answers](https://www.sourcesignal.ai/docs/spot-and-agents/how-the-door-answers.md): What your own agent meets when it calls sourcesignal over MCP: one address, the token on every request, the limits on each call, long answers in parts and slow work it checks on later.
- [Connect your own agent over MCP](https://www.sourcesignal.ai/docs/spot-and-agents/connect-your-own-agent-over-mcp.md): Give Claude Code, Claude Desktop or any other MCP client a token of its own, and it works in sourcesignal as you, with only the scopes and projects you choose. Looking for an API? The agent connection over MCP is on every plan, with a token from Settings, Agents; a custom API for your own systems is an Enterprise conversation.
- [Agent scopes and limits](https://www.sourcesignal.ai/docs/spot-and-agents/agent-scopes-and-limits.md): Your agent works as you, with your plan's limits and no more. Every change it prepares waits for you, unless your setting under Settings, Agents applies that kind of change at once or a mandate you approved lets it make that change on its own.
- [Change cards: apply and undo](https://www.sourcesignal.ai/docs/spot-and-agents/change-cards-apply-and-undo.md): Every change Spot or your own agent prepares reaches you as a change card, unless your setting under Settings, Agents applies that kind of change at once. A card changes nothing in your project until you press Apply, and many changes can be undone for 10 minutes.