Feature

Agent API and MCP server for LinkedIn content

The agent API in ContentOS lets AI agents and your own scripts use the product directly. An MCP server connects assistants such as Claude, Cursor and Codex, and a REST API with scoped keys covers drafting, posts, publishing, the engagement inbox and warm contacts.

Updated

What can the MCP server do?

The ContentOS MCP server gives an assistant tools to search your posts, list the calendar, read an analytics summary and post performance, create and publish posts, generate drafts and manage Autopilot accounts. Claude connects with OAuth; other clients use an API key.

Settings has an Agents and API tab with one-click install for Cursor, VS Code, Claude, Claude Code and Codex, so setup is a copy or a click rather than editing config files by hand.

What does the REST API cover?

The agent REST API covers sessions, drafting, posts, publishing, the engagement inbox (sync, drafted replies) and warm contacts. A skill document describes the endpoints in Markdown for agents to read, and an OpenAPI description is published for tooling.

  • Keys are created in Settings and shown once.
  • Scopes: read, write, publish and outreach.
  • Requests authenticate with a bearer key.

What limits apply to agents?

Agents act within your plan. Publishing through the API follows the same publishing rules as the app, and Autopilot tools need a plan that includes Autopilot. Give each key only the scopes its job needs, and revoke keys you no longer use.

What do people build with the agent API?

Common uses are an assistant that drafts next week's posts from your notes and queues them for review, a script that pulls performance summaries into a weekly report, and an agent that checks the engagement inbox and prepares drafted replies for you to approve. Because the API covers drafting, scheduling, publishing and the inbox, an agent can run the same loop you run in the app.

The MCP server makes this possible without writing code: install it in Claude, Cursor or Codex, then ask in plain language for the calendar, a draft or a performance summary.

Start with read-only keys while you experiment, so an assistant can look at your calendar and analytics without changing anything. Add the write and publish scopes only once you trust the workflow, and keep separate keys for separate tools so you can revoke one without touching the others.

Which plans include agent api and mcp server?

Availability by plan:

  • API keys and the MCP server are available to signed-in accounts.
  • Publishing through the API needs a paid ContentOS subscription; Autopilot tools can run on every connected account.

Frequently asked questions

Which AI assistants can connect to ContentOS?

Any MCP client. Settings offers one-click install for Cursor, VS Code, Claude, Claude Code and Codex. Claude connects with OAuth, and other clients use an API key you create in Settings.

Can an agent publish posts for me?

Yes, if its key has the publish scope and your plan allows publishing. Publishing through the API follows the same rules as the app. Use separate keys with narrower scopes for agents that only need to read or draft.

Is there documentation for the API?

Yes. The API publishes a Markdown skill document for agents and an OpenAPI description for tooling. The llms.txt file at the site root also points agents to the main resources.

Do I need to write code to use the agent API?

No. The MCP server works inside assistants such as Claude, Cursor and Codex with a one-click install from Settings, and you talk to it in plain language. The REST API is there for developers who want to script ContentOS from their own tools.

Try agent api and mcp server on your own posts

Start a free trial, connect LinkedIn, and see the feature work on your real content.