Pythe

Connect your AI app

Step-by-step setup for Claude, ChatGPT, Cursor, Claude Code and Codex, plus the REST fallback for tools that cannot sign in with OAuth.

Updated 2026-10-13

The server URL is https://pythe.app/mcp. In Pythe, the Deploy page (/connect-agent) has these steps for each client with copy buttons, a task picker and ready-made prompts for the company you have selected.

Claude (web)

  1. Open custom connectors. In Claude, go to Settings, Connectors, Add custom connector. (This is different from Claude Code’s terminal setup, and availability depends on the account.)
  2. Add Pythe. Name it Pythe and paste the server URL. Use OAuth.
    https://pythe.app/mcp
  3. Connect your workspace. Choose Add, sign in to Pythe, pick the workspace Claude may use, and allow the connection.
  4. Use it. Enable Pythe in a new chat and ask: “Use Pythe to show my operating plan.”

ChatGPT

  1. Open the custom MCP form. In Settings, Apps or Connectors, Advanced settings, enable Developer mode, then create or connect a custom MCP.
  2. Choose Streamable HTTP. Name it Pythe and select Streamable HTTP with OAuth. Do not enter a command, arguments or environment variables: Pythe is remote.
  3. Paste the URL. Leave Client ID and Client Secret empty.
    https://pythe.app/mcp
  4. Connect your workspace. Save, sign in to Pythe, select the workspace and allow.
  5. Use it. Enable Pythe in the chat and ask: “Use Pythe to show my operating plan.”

Cursor

  1. Open the MCP configuration. Cursor Settings, Tools & MCP, New MCP Server. Use ~/.cursor/mcp.json for every project or .cursor/mcp.json in a project for project-only access.
  2. Add the remote server. Merge this into your configuration. No API key or Authorization header is needed.
    {
  "mcpServers": {
    "pythe": {
      "url": "https://pythe.app/mcp"
    }
  }
}
  3. Authorize. Select Connect next to Pythe, complete OAuth in the browser and choose the workspace.
  4. Verify. Start a new Agent chat and ask: “Use Pythe to show my current priority.” A green server status and a context reply confirm setup.

Claude Code

  1. Register Pythe. Run this in a terminal. The default scope is the current project. Add --scope user before the server name if you want it in every project.
    claude mcp add --transport http pythe "https://pythe.app/mcp"
  2. Authorize. Open /mcp inside Claude Code, select Pythe, complete OAuth in the browser and choose the workspace.
  3. Verify. Confirm Pythe reports connected, then ask for your current priority.
    claude mcp list
  4. If it fails. Restart Claude Code in case the session cached old discovery. To start over, remove it and add it again.
    claude mcp remove pythe

Codex

  1. Add the remote server. This writes the Streamable HTTP entry to Codex’s config.
    codex mcp add pythe --url "https://pythe.app/mcp"
  2. Authorize. Start the OAuth flow, sign in to Pythe and choose the workspace.
    codex mcp login pythe
  3. Restart and verify. Restart the Codex client, then confirm Pythe is connected.
    codex mcp list

Other clients

Any client that supports remote MCP with OAuth works with the same URL. For tools that cannot complete OAuth (the Deploy page lists OpenClaw and Hermes Agent as examples), use the REST gateway with a scoped key.

Give your client a job

After connecting, the Deploy page can generate a prompt for your company. It confirms and switches to the exact workspace first, then calls get_task_prompt at run time, so a stale label in the page can never be the authoritative brief. Example you can paste into Claude Code in the repository you want to work in:

Example prompt
Read my live Pythe workspace and call get_task_prompt with goal "build" and mission "THE_MILESTONE_ID". Implement the brief in this repository, follow project instructions, run relevant checks, and save the complete change report with save_note attached to that mission. Return the Pythe link. Ask me before publishing or marking the milestone complete.

Related

·