# MCP Server > The Mailrith Connector is available in Claude, and the same remote MCP server supports OAuth or protected server-side credentials for OpenAI, n8n, Pipedream, and other clients. - Human page: https://mailrith.com/developers/mcp - Markdown page: https://mailrith.com/developers/mcp.md - Category: AI Agent Platform - Reading time: 7 min read - Last updated: 2026-08-25 - Related keywords: MCP Server, MCP Server developer docs, Mailrith MCP server, Mailrith Claude Connector, Claude Connectors Directory, remote MCP email marketing server, AI Agent Platform, AI Agent Platform developer docs, Mailrith developer docs, Mailrith public API, Official MCP Server, OpenAI Responses API, Claude Connector And Messages API, n8n and Pipedream, Agents, SDKs, API Reference ## AI Agent Notes - Use this page as implementation guidance, then validate exact endpoint fields against the OpenAPI document. - Keep API keys server-side and workspace-scoped unless a guide explicitly says otherwise. - Do not invent privacy, consent, or lawful-basis evidence. Send only fields that appear in the OpenAPI schema for the endpoint you are using. ## What This Guide Covers Connect Mailrith from Claude's Connectors Directory or use the official remote MCP server with other supported agent runtimes. ## Official MCP Server Mailrith provides an official remote MCP endpoint for agent runtimes that work better with tools than with raw HTTP calls. The MCP server exposes the public API as named tools that agents can inspect and call. The remote endpoint is `/mcp` on the API origin. It accepts workspace API keys and OAuth access tokens, so workspace scope and permissions behave the same way as the REST API. Use the MCP server when the agent platform supports MCP directly. Use the REST API or SDKs when you need a standard backend integration, custom retry behavior, or full control over how requests are built. Every client uses the same server URL and sees the same reviewed focused-tool catalog. During OAuth, choose what the connection should do. Mailrith grants the matching permissions and checks them again when each tool runs. 1. Choose OAuth for user-connected runtimes such as Claude Desktop. Create a workspace API key only for server-side runtimes that can safely store bearer credentials. 2. Decide which Mailrith tools the agent is allowed to use. Keep the allowed tool list as small as possible. 3. Configure the agent runtime to use `https://api.mailrith.com/mcp`. If the runtime supports custom authorization headers, set `Authorization: Bearer `. 4. For OAuth, choose the Work Profile that matches the job on Mailrith's approval screen. Start with Reporting when the workflow only needs reads. If a later task needs another resource, reconnect from the agent app and choose the matching Work Profile. 5. Test the workflow with a small prompt. Then check Mailrith and confirm the agent made only the expected actions. 6. Add write tools only after the read flow works reliably. 7. Grant only the resource permissions the connected agent needs. Mailrith checks the resource permission and workspace on every request, without adding a second action-approval step. - Preferred transport: Streamable HTTP. - Authentication: OAuth for Claude Desktop and other user-connected clients; `Authorization: Bearer ` for code-first clients. - The hosted server exposes a fixed reviewed catalog of focused tools such as `subscribers_list`, `broadcasts_create`, `broadcasts_preflight`, and `broadcasts_send`. - Limit the tools exposed to an agent whenever the platform supports tool limits. A narrow toolset is easier to understand and safer to approve. - Every submitted tool includes its exact input and output schema, OAuth permissions, and safety annotations. ## OpenAI Responses API OpenAI's MCP tool integration can connect directly to Mailrith's remote MCP server and use focused tools with exact schemas. Use `allowed_tools` to keep only the focused tools needed for the job. A reporting workflow can keep capability, workspace, and Subscriber reads. A Broadcast preparation workflow can add sender, targeting, Template, draft, and preflight tools without enabling schedule or send. Store the Mailrith API key in a server-side secret, and pass the key as an authorization header in the tool configuration. **Use Mailrith as an MCP Tool Source in OpenAI** ```ts import OpenAI from "openai"; const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY, }); const response = await openai.responses.create({ model: "gpt-5.1", input: "List the newest subscribers and draft a re-engagement broadcast.", tools: [ { type: "mcp", server_label: "mailrith", server_url: "https://api.mailrith.com/mcp", headers: { Authorization: `Bearer ${process.env.MAILRITH_API_KEY}`, }, allowed_tools: [ "discovery_get_capabilities", "subscribers_list", "sender_identities_list", "segments_list", "email_templates_list", "broadcasts_create", "broadcasts_preflight", ], require_approval: { always: { read_only: false, }, }, }, ], }); ``` ## Claude Connector And Messages API The Mailrith Connector is available in Claude's Connectors Directory. It uses OAuth, so users do not paste workspace API keys into Claude. Code-first integrations that use the Claude Messages API can connect to the same remote MCP server with a task-specific Mailrith API key stored in a server-side secret manager. On the Mailrith approval screen, check the requested permissions before approving. Mailrith grants the permissions requested by the connected app, so configure the connector or client to request only the scopes needed for its workflow. Configure the connector with only the tools needed by the Claude workflow. Do not allow write or send tools for read-only workflows. When troubleshooting permission errors, disconnect Mailrith and connect it again. Then confirm the connection requested the permissions required for the workflow. For example, drafting or activating a Sequence requires `sequences:write`, and inspecting templates requires `email_templates:read`. 1. Open Claude or Claude Desktop, then open `Customize` > `Connectors`. 2. Select the `+` button, choose `Browse connectors`, and search for `Mailrith`. 3. Choose `Mailrith`, review its capabilities, and select `Connect`. 4. Sign in to Mailrith, choose one workspace, review the requested permissions, and select `Allow Access`. 5. Begin with a read-only request. Add write tools only when the task needs them, then confirm every change from the saved Mailrith resource or its progress endpoint. **Use Mailrith With The Claude Messages API** ```ts const anthropicApiKey = process.env.ANTHROPIC_API_KEY; const mailrithApiKey = process.env.MAILRITH_API_KEY; if (!anthropicApiKey || !mailrithApiKey) { throw new Error( "Set ANTHROPIC_API_KEY and MAILRITH_API_KEY in the server's secret manager.", ); } const response = await fetch("https://api.anthropic.com/v1/messages", { method: "POST", headers: { "content-type": "application/json", "x-api-key": anthropicApiKey, "anthropic-version": "2023-06-01", "anthropic-beta": "mcp-client-2025-11-20", }, body: JSON.stringify({ model: "claude-sonnet-4-6", max_tokens: 1200, messages: [ { role: "user", content: "List the newest subscribers and propose a broadcast.", }, ], mcp_servers: [ { type: "url", url: "https://api.mailrith.com/mcp", name: "mailrith", authorization_token: mailrithApiKey, }, ], tools: [ { type: "mcp_toolset", mcp_server_name: "mailrith", default_config: { enabled: false }, configs: { discovery_get_capabilities: { enabled: true }, subscribers_list: { enabled: true }, sender_identities_list: { enabled: true }, segments_list: { enabled: true }, email_templates_list: { enabled: true }, broadcasts_create: { enabled: true }, broadcasts_preflight: { enabled: true }, }, }, ], }), }); if (!response.ok) { throw new Error("Claude request failed with status " + response.status); } const result = await response.json(); // Pass result only to the authorized request handler. // Never write Subscriber or Broadcast content to application logs. ``` ## n8n and Pipedream Workflow platforms usually provide MCP connection fields in a UI instead of a code-first SDK. Mailrith's remote server works with these clients by using a streamable HTTP URL and a Bearer token. Store the Mailrith key in environment variables or in the platform's secret manager. Do not paste the raw key into step names, logs, shared screenshots, or exported workflow files. Start with one low-risk tool, such as listing subscribers or reading workspace capabilities. Add write tools only after you confirm the workflow performs the expected actions. **n8n MCP Client Fields** ```text Transport: Streamable HTTP Server URL: https://api.mailrith.com/mcp Authentication: Header Header name: Authorization Header value: Bearer {{$env.MAILRITH_API_KEY}} ``` **Pipedream Remote MCP Server Fields** ```text Server URL: https://api.mailrith.com/mcp Authorization header: Bearer {{process.env.MAILRITH_API_KEY}} Suggested tools: discovery_get_capabilities, workspace_get, subscribers_list, subscribers_get ``` ## Related Guides - [Agents](https://mailrith.com/developers/agents.md): Mailrith supports OAuth directory connections for ChatGPT, Codex, and Claude, plus a compact discovery stack with llms files, API metadata, OpenAPI, authenticated capabilities, SDKs, and MCP entry points. - [SDKs](https://mailrith.com/developers/sdks.md): Mailrith ships official TypeScript and Python SDKs generated from the same public contract that powers the OpenAPI document and capability map. - [API Reference](https://mailrith.com/developers/api-reference.md): The full API reference is generated from the same public contract used by the API worker and SDK tooling. Use the API reference to find exact paths, methods, parameters, request schemas, response schemas, operation IDs, and the downloadable OpenAPI document.