Everything needed to drive planpo.st from an AI agent or your own code: scheduling, publishing, video and carousel rendering, approvals, and revenue attribution.
In the app, go to Settings, then API Keys, then Create Key. Pick only the permissions the agent needs. Keys are shown once and stored hashed, so copy it before closing the dialog.
The hosted endpoint needs nothing installed. This is Claude Code; every other client is on its own page below.
claude mcp add --transport http planpost https://mcp.planpo.st/mcp \
--header "Authorization: Bearer pk_live_xxx"Nothing else to configure. Ask in plain language.
Schedule this to LinkedIn and X for tomorrow 9am: ...
Make a quote video from this testimonial and publish it as a reel.
Which posts made money last month?Config format and file location differ per client. These were checked against each vendor's own documentation.
Anthropic's terminal and IDE agent. The quickest client to set up.
The desktop app. Supports both a config file and custom connectors.
The AI code editor. Config is per project or global.
OpenAI's terminal agent. Configured in TOML rather than JSON.
Google's terminal agent. Uses httpUrl for remote servers.
Connects through custom connectors. Availability depends on your ChatGPT plan and whether developer mode is enabled.
Google's agentic IDE and CLI. Since 2.x the IDE, the agy CLI and the SDK read one shared MCP config.
Nous Research's open-source agent, a desktop app and CLI that also runs in Telegram, Discord, Slack, WhatsApp, Signal and email.
The server is a standard MCP implementation, so anything that speaks the protocol will work.
posts:read cannot publish. Grant the narrowest set that works.