ReadMe and TextMine MCP Quickstart

Connect Codex, Claude, Cursor, Windsurf, ChatGPT, or another MCP-compatible client to TextMine documentation and tools.

TextMine currently exposes two MCP surfaces. They are related, but they solve different jobs.

MCP endpoints

Use the environment-specific TextMine host for dedicated deployments. The public multitenant host is https://app.textmine.com.

When to use each MCP

• Use ReadMe MCP when the agent needs to search API docs, inspect schemas, or discover endpoint details from developers.textmine.com.
• Use TextMine-hosted MCP when the agent needs to operate TextMine resources directly through the hosted product tool surface.
• Use Public API V3 directly when the client already has an HTTP integration layer and does not need MCP tool discovery. Self-service agents can complete signup and setup entirely through REST.
• Use Agent self-service signup first when an autonomous third-party agent needs its own TextMine organization and service-account API key.

OpenAPI and MCP

Use the published API reference at developers.textmine.com/reference. Agents and developers use it for endpoint discovery, SDK generation, typed clients, API-client imports, and exact request/response shapes.

MCP clients do not use OpenAPI as their runtime contract. They initialize a Streamable HTTP session, list tools, and call tools using MCP schemas. Some platforms, including Copilot Studio custom connectors, may need an OpenAPI wrapper around /mcp; that wrapper is platform adapter metadata, not the canonical Public API spec.

Authentication

Public API V3 and the TextMine-hosted MCP endpoint require a TextMine bearer key:

Authorization: Bearer tm_your_api_key

Platforms that own the Authorization header can use the alternate API-key header:

X-TextMine-Api-Key: tm_your_api_key

Prefer bearer authentication for direct Public API and MCP clients. Create or request a key through TextMine support, an existing organization admin flow, or Agent self-service signup. Store the key in a trusted runtime or secret manager. Do not expose it in prompts, chat messages, task transcripts, or model-visible context.

Generic remote MCP config

Use your MCP client's remote-server syntax. This example shows the two surfaces side by side:

{
  "mcpServers": {
    "textmine-docs": {
      "url": "https://developers.textmine.com/mcp"
    },
    "textmine": {
      "url": "https://app.textmine.com/mcp",
      "headers": {
        "Authorization": "Bearer tm_your_api_key"
      }
    }
  }
}

Some clients store headers separately from the MCP server JSON. Follow the client's secret-storage pattern whenever possible.

Streamable HTTP handshake

The TextMine-hosted MCP endpoint uses Streamable HTTP. Direct clients must:

1. Send initialize.
2. Keep the returned mcp-session-id response header.
3. Send notifications/initialized with mcp-session-id.
4. Call tools/list or tools/call with mcp-session-id.

A direct tools/call without mcp-session-id is rejected.

Agent self-service signup

Self-service agents do not need MCP for signup. For multitenant, non-enterprise onboarding, use Public API V3:

• GET https://public-api.textmine.com/.well-known/textmine-agent.json
• GET https://public-api.textmine.com/v3/third-party-agents/signup/discovery
• GET https://public-api.textmine.com/v3/third-party-agents/signup/availability
• POST https://public-api.textmine.com/v3/third-party-agents/signup/self-service

The signup response creates a TextMine organization, service-account user, encrypted tm_... API key, and billing account. With encrypted credential delivery, the trusted runtime decrypts encryptedApiKey and stores it outside model context. After signup, use REST for Vault creation, document type creation, file upload, tasks/messages, and billing status unless the agent platform specifically wants MCP.

First useful checks

1. Ask ReadMe MCP for the current Public API V3 docs for Vaults, documents, tags, records, workflows, and agent signup.
2. Verify the TextMine API key by calling a low-risk endpoint such as the current user/profile or Vault list endpoint.
3. List Vaults before Vault-scoped actions.
4. For document workflows, upload/import, poll processing status, then fetch tags, metadata, extracted text, or tag source chunks as needed.
5. For billing-gated agent organizations, call GET /v3/agent-billing/status before heavy work.