Skip to main content
The MCP server has one tool per API route, named by its operation id. A connected agent can do what you can do in the dashboard with your role, except create API keys.

​
Read where you stand

​
Read the Plan

list_actions returns the Plan, best first, with counts per status. get_action returns everything needed to do one action: its evidence, the prompts it should move, the target page, the change Jam prepared (if any) and agentPrompt, instructions your agent can follow, ending with how to report the work done. See The Plan.

​
Run research

get_research says whether research is running and whether another run would be accepted. get_research_report returns what the latest run found. create_research_request asks Jam to research again and rewrite the Plan: the first run is free, each later one holds 500 credits, kept only if it succeeds. It takes about ten minutes. See Research. create_visibility_run asks every prompt again now. Scans use no credits; it answers 409 while a scan is already running.

​
Ship work

An agent can do an action three ways:
  • Have Jam do it. create_action_execution drafts a new page, or prepares a code change for a page edit or technical fix. With {"mode": "agent"} it starts the same background run as the dashboard’s Send to Jam.
  • Do it itself. Follow the action’s agentPrompt. For a new page: read the brief (get_opportunity_brief), claim it (claim_opportunity), write it, and store the draft (create_content).
  • Ship it. Open the pull request for a prepared change (create_site_change_pull_request, 20 credits, kept only if it opens), or publish a draft to your connected repository (publish_content).
Then it reports the work with complete_action and the URL of what changed. Jam checks the page and marks the action verified or unverified. See Doing the work and Delivery.

​
Manage what is tracked

Add, reword, pause, replace or remove prompts (add_prompt, add_prompts_bulk, edit_prompt, pause_prompt, replace_prompt, remove_prompt), ask Jam for more (suggest_prompts), and edit the brand and competitors (update_brand, add_brand_competitors, remove_brand_competitor).

​
Get told when things happen

Send Jam’s notifications to your own URL with create_webhook, and choose which ones with update_notification_rule. See Webhooks.

​
How tools are marked

  • Reads are readOnlyHint: true.
  • Tools that spend credits, publish to your site or repository, or delete something are destructiveHint: true. A tool that spends credits says what it costs at the start of its description.
  • Tools that are safe to repeat are idempotentHint: true.
Set your client to approve reads on its own and ask before the rest. Writes need an active plan; when a plan ends the workspace becomes read only and writes answer 402.

​
A first conversation