Connect Claude to Mastodon

Can Claude post to Mastodon?

Postgale is the secure layer for AI agents on social media. MCP (Model Context Protocol) is the open standard that lets AI assistants such as Claude and Cursor use outside tools. An MCP server is what they connect to. Claude does not reach Mastodon on its own; with Postgale connected it gets tools for the Mastodon accounts you connected and nothing else.

What do you need?

• A Mastodon account on any server (mastodon.social, hachyderm.io, your own instance...).
• A Postgale workspace, and a sign-in or an API key for the agent.
• Optional images or a video on a public HTTPS URL (upload them with upload_media).
• Claude, set up as below (claude.ai, Claude Desktop, Claude mobile and Claude Code).

Step 1: connect your Mastodon account

1. Create a Postgale workspace and open Dashboard → Accounts → Mastodon.
2. Enter your server’s domain (for @[email protected] that is hachyderm.io), sign in there and approve Postgale.
3. Connect your AI tool: add https://postgale.com/mcp in ChatGPT, Claude or Cursor and sign in to Postgale, or create a key under Dashboard → API keys for n8n or your own code.
4. Ask the agent for a draft first. Turn on approval in the dashboard if a person should sign off before anything goes live.

An agent can also start the connection itself: the connect_account tool (or GET /api/v1/connect/mastodon?instance=your.server) returns a sign-in link that is valid for 10 minutes and that the account owner opens in a browser. The agent never sees the password or the token.

Step 2: connect Claude to Postgale

1. Create a Postgale workspace and connect the social accounts the agent may use under Dashboard → Accounts.
2. In Claude, open Settings → Connectors → Add custom connector, and paste the server URL:

https://postgale.com/mcp

1. Choose Connect. A Postgale page opens in the browser: sign in, pick the workspace the app may use and approve. The consent page names the app and where it returns to; approve only requests you started yourself.
2. Back in Claude, enable the Postgale tools in a new chat and ask: “Which social accounts are connected in Postgale?” Claude calls list_accounts and answers with the accounts and their status.

How do I set up Claude Code?

1. Create a Postgale workspace and connect the social accounts the agent may use under Dashboard → Accounts.
2. In a terminal, add the server:

claude mcp add --transport http postgale https://postgale.com/mcp

1. Run /mcp inside Claude Code, select postgale and choose Authenticate. A Postgale page opens in the browser: sign in, pick the workspace the app may use and approve. The consent page names the app and where it returns to; approve only requests you started yourself.
2. Ask: “Which social accounts are connected in Postgale?” to check the connection.

To use an API key instead (for example on a server without a browser), create one under Dashboard → API keys and add the header:

claude mcp add --transport http postgale https://postgale.com/mcp \
  --header "Authorization: Bearer sk_…"

What can I ask Claude to do on Mastodon?

Check the connection first: “Which Mastodon accounts are connected in Postgale?” Then, in plain language:

Ask your agent“Post this release note to our Mastodon account with the content warning “release notes”.”

Ask your agent“Write alt text for this image and post it to Mastodon with the caption from the brief.”

Ask your agent“Turn these notes into a short Mastodon thread, as drafts.”

{
  "content": "Release 2.4 is out: faster sync and dark mode. https://example.com/changelog",
  "platforms": [
    {
      "accountId": "acc_masto",
      "platformSpecificData": {
        "mediaUrl": "https://media.example.com/screenshot.png",
        "altText": "The new dashboard in dark mode",
        "visibility": "unlisted",
        "language": "en"
      }
    }
  ],
  "publishNow": true
}

Which Mastodon rules does Postgale enforce?

• Text is counted the way your server counts it (links as 23 characters, remote mentions by username, content warnings included) against your server’s limit, usually 500.
• Media is uploaded with its alt text and the status waits until your server has processed it.
• Every status carries an idempotency key, so a retried request never posts twice.
• If you revoke Postgale under Authorized apps, the account is marked as needing reconnection and your webhook hears about it.
• When your server asks to slow down, Postgale waits until its limit resets before retrying.

How do I keep a person in the loop?

• Approval: tick “Hold posts from this app for approval” when you sign in, or in the Postgale dashboard require approval for the API key the agent uses (or for the whole workspace, or one account). Posts then wait as pending approval; comment replies and moderation, direct messages, WhatsApp sends and edits to live posts wait in the same queue. A person approves, edits or rejects each one in Dashboard → Approvals; the agent’s own key or sign-in cannot approve.
• Tool annotations: every Postgale tool declares whether it only reads, writes, or publishes live, deletes or spends. Clients that honour annotations ask before destructive tools run.
• Audit log: every write the agent makes is recorded with the key or the connected app, the action and the time.
• Revoking: each sign-in appears under Dashboard → Connected apps and each key under Dashboard → API keys; either can be revoked at any time.

What can’t an agent do on Mastodon through Postgale?

• Edit a status (not supported yet); delete and post again instead.
• Read favourites, boosts or replies yet.
• Create polls.

Common questions