Model Context Protocol

One MCP server.
Every social platform.

PostLake ships a hosted MCP server, so any MCP-capable agent, Claude, ChatGPT, Cursor and more, can publish, schedule and analyse across every network just by being asked. No SDKs, no glue code, no keys pasted into a chat.

Get started free →

Server URL

https://api.postlake.dev/mcp

Connect in three steps

It takes about three minutes, and not a single line of code.

Add the server

In your agent's connector settings, add a new MCP server and paste the URL above. In Claude Code it's one command: claude mcp add --transport http postlake …

Approve once

You're sent to the PostLake Authorize screen. Sign in and click Approve. OAuth means you do not need to paste an existing API key into the client.

Just ask

Tell your agent what to do in plain English, "post this to LinkedIn and X, schedule a recap for Friday." It runs the tools for you.

A full toolset your agent can run

The full social workflow is exposed as MCP tools. Lists and actions return compact markdown, while complete records return structured JSON when every field matters.

Publish & schedule

Upload local files without public hosting, import remote media, post now or later, and edit, cancel or retract posts. create_post, edit_post, cancel_post, upload_media, upload_media_batch, validate_post

Read & respond

Read comments and notifications, reply, and engage with posts or people across supported networks. list_notifications, read_comments, engage

Manage profiles & channels

Create profiles, inspect destinations, connect supported channels, or give the owner a signed approval link. list_social_accounts, get_connect_link

Direct messages

List conversations, read a thread, reply, or start a supported conversation from a handle.

Read analytics

Per-post and cross-platform performance, so the agent can decide what to post next. get_analytics, get_post_analytics

Safe by default

Idempotency keys mean a retrying agent never double-posts, even at 3am.

Platform-aware

Live limits and creator-level constraints are checked before posting, so the agent only offers valid options. get_platform_capabilities, get_publish_info

Secure by design

OAuth connectionYou approve each agent on a PostLake screen without pasting an existing API key into the client. Streamable HTTP hosts such as Cursor get HTTP 401 on first connect so they can show Connect; directories may still probe initialize without a token.

Revoke anytimeEvery connected agent is listed under Agents in your dashboard. One click cuts off its access instantly.

Scoped to youAn agent acts only on your account, through the same permission checks and rate limits as the API.

Works with your agent

Any MCP-capable client. Step-by-step setup for the popular ones:

Claude → ChatGPT → Cursor → Gemini → GitHub Copilot →

Plain markdown for agents · llms.txt

Common questions

Do I need an API key to connect my AI agent?

No. The MCP connection uses OAuth, so you do not paste an existing API key into the client. You can revoke any agent's access from Agents at any time.

Which AI assistants work with PostLake's MCP server?

Any MCP-capable client. Claude (Desktop and Code), ChatGPT, Cursor, Gemini CLI, VS Code / GitHub Copilot, and any other host that can add an MCP server. Point it at https://api.postlake.dev/mcp.

What can my agent do once connected?

Publish, schedule and validate posts; upload media; read and respond to activity; manage profiles and channels; read analytics; and create a full-access API key for another trusted service when asked.

Is it safe to let an AI agent post on my behalf?

Only approve clients you trust. Idempotency keys prevent duplicate posts, publishing limits are enforced, and every connected agent is listed under Agents where you can revoke access instantly.

How much does the MCP server cost?

The MCP server is included on every plan, free tier too. You pay usage-based credits per post. One credit per post (X costs more), with a free tier of 20 credits a month and no card.