Skip to content

MCP server

Chat apps such as Claude and ChatGPT, and coding agents such as Claude Code and Codex, support the Model Context Protocol (MCP). Any of them can work in your workspace through our MCP server:

https://mcp.reminix.com/mcp

Add the address above to your agent as a remote MCP server:

  • Claude: Settings → Connectors → Add custom connector.
  • ChatGPT: add a custom connector in ChatGPT’s settings.
  • Claude Code: run claude mcp add --transport http reminix https://mcp.reminix.com/mcp, then /mcp in Claude Code to sign in.
  • Codex: run codex mcp add reminix --url https://mcp.reminix.com/mcp, then codex mcp login reminix.
  • Cursor, GitHub Copilot in VS Code, Windsurf, Gemini CLI, or another app that supports remote MCP servers: add the address in its MCP settings, and sign in when it asks.

The first time the agent connects, your browser opens:

  1. Sign in, if you aren’t already.
  2. Choose the one workspace the agent will work in.
  3. Choose what the agent may do. Anything that can change data starts unticked, so tick only what the agent needs.
  4. Click Allow.

There’s nothing to copy and no key to store. The agent gets its own sign-in, which works only for the workspace and the access you chose.

The agent’s tools are the API operations you allowed, such as reading the workspace or listing webhook deliveries. It always acts as you, with your role in the workspace, so it can’t do anything you couldn’t do in the app. Everything it changes shows up in the workspace’s audit log under your name.

Each published script you may run is also a tool the agent calls by name. A script called refund-customer becomes a tool called refund_customer. The tool takes the script’s inputs, and its description starts with the script’s own.

If two or more environments allow agents, each script’s tool also takes an optional environment argument, such as staging. If the script has its own input called environment, the argument’s name is reminix_environment instead. A run uses the first of these that’s set:

  1. The tool call’s environment argument.
  2. The environment in the server address, such as https://mcp.reminix.com/mcp?environment=staging.
  3. The environment you picked when you connected the agent. The sign-in screen offers the environments the agent may use, and starts on the workspace’s default.

Every tool’s response ends with the environment it acted in. Agents can list environments, but they can’t add, change or delete them. An environment can also keep agents out entirely.

An agent can rename a script. It can’t change who may run the script or which of its runs need approval.

Even with full access, an agent can’t send your workspace’s data somewhere new or hand out access by itself, for example by creating a webhook endpoint or a credential. It can ask instead, and you approve or deny the request in the app. If the action creates a secret, the app shows it to you, never to the agent. Approvals explains how.

  • Account → Connected apps lists every agent you’ve connected. Disconnect one and it stops working at once.
  • The workspace’s audit log records what each agent did.
  • If you’re removed from the workspace, your agents lose access too.

The server follows the MCP authorization specification. A request without a token gets a 401 that points to /.well-known/oauth-protected-resource/mcp, which names the authorization server, https://api.reminix.com/auth. That server supports dynamic client registration and PKCE, and issues tokens for the MCP server only.

Every tool that changes something takes an optional idempotencyKey. If an agent calls the tool again after an error or a timeout, the same key makes sure the change happens only once. Within a single call, the server also retries a failed attempt with that call’s key. Idempotency has the details.

For your own scripts, the CLI or the SDK is simpler. For the agent skill, see Build with AI agents.