Skip to main content
The TextMine MCP server is the TextMine-hosted Streamable HTTP MCP surface for external agent clients. It lets Codex, Claude, Cursor, Windsurf, ChatGPT, Microsoft Copilot, and automation clients discover and call TextMine tools through an agent-friendly protocol. Use it when an external agent needs to read or act on TextMine data. Use Workbench when a human is working inside the TextMine app.

​
Public endpoint

Use the central TextMine MCP endpoint for public setup guides, MCP registry metadata, and marketplace-style configuration:
Use a different TextMine-issued MCP host only when a customer-specific deployment guide explicitly provides one. Do not use https://developers.textmine.com/mcp for TextMine product tool calls. That is the developer hub MCP endpoint for API reference discovery.

​
How it fits

TextMine has three agent-facing surfaces that are easy to confuse:
  • TextMine MCP server exposes TextMine product tools to external MCP clients. It is backed by Public API V3.
  • ReadMe MCP server is hosted by the developer hub and helps agents discover API documentation and generated API operations from developers.textmine.com.
  • Workbench MCP providers are in-app integrations that let Workbench browse and import files from connected providers such as storage systems.
For API operation details, examples, and client setup, use API and developer docs or the developer hub’s agent guide. For provider and integration inventory, see Integration catalog.

​
Set up a client

Configure MCP clients with:
MCP authentication has two supported patterns: Direct MCP clients that can send an Authorization header can use a TextMine API key as a bearer token:
OAuth-enabled app flows can also send an OAuth access token as the bearer token:
Use the OAuth flow described by the app or marketplace setup guide. TextMine resolves the user, organization, and environment server-side; users should not paste tm_... keys or OAuth tokens into prompts, app package manifests, screenshots, or shared documentation. Connector platforms that cannot set an Authorization header can send the TextMine key in the alternate header:
OAuth client registrations can use TextMine’s protected-resource metadata:
Streamable HTTP clients must create a session before calling tools:
  1. Send initialize to https://mcp.textmine.com/mcp.
  2. Keep the returned mcp-session-id response header.
  3. Send notifications/initialized with that mcp-session-id.
  4. Call tools/list or tools/call with the same mcp-session-id.
A direct tools/call without mcp-session-id is rejected.

​
What it exposes

The MCP server is implemented as a thin wrapper over Public API V3. Each MCP tool should map to an existing V3 endpoint rather than adding new business logic in the MCP service. Current tool areas include:
  • Vaults and Vault documents.
  • Document search, create, update, and delete.
  • Reports and report templates.
  • Workflows and workflow catalogs.
  • Agents.
  • Alerts.
  • Team and Vault users.
  • Integrations, provider connection status, provider folder listing, and provider file listing.
When an agent needs exact request or response shapes, route it to the Public API reference. When it needs product semantics, route it back to this Mintlify docs site.

​
Import and provider support

TextMine MCP clients can use provider-facing tools where the connected provider supports them. That includes listing integrations, checking provider connection status, listing provider folders or files, and uploading a provider file into Vault through TextMine. This is separate from Workbench’s in-app provider import, but both use the same product concept: external provider files can become TextMine context or TextMine Vault documents. Use this split when documenting or building agents:
  • Import into Workbench when the file is temporary context for chat, comparison, summarization, or drafting.
  • Import into Vault when the file should be processed, extracted, reviewed, governed, used by workflows, or published onward to Legislate.
  • Use Public API V3 when an external system or agent needs a durable integration point.

​
Authentication model

The MCP service forwards the authenticated caller to Public API V3. Public API resolves the real user or service-account organization and applies normal TextMine permissions. The caller can be represented by a tm_... TextMine API key or, for enabled marketplace/shared-host app flows, an OAuth bearer token. OAuth callers are resolved server-side before the MCP request reaches Public API V3. Self-service agents should first get a tm_... service-account key through Agent self-service signup. Enterprise, private, and customer-dedicated deployments should use support-led provisioning unless their setup guide provides an OAuth connection flow. Do not rely on a shared service user for user-facing MCP clients. Each connection should represent the TextMine user or service account whose Vaults, records, integrations, and tasks the agent may access. For setup examples, see the developer hub pages for TextMine MCP, Codex, Claude, Cursor and Windsurf, and ChatGPT/OpenAI. For command-line workflows, use TextMine CLI and CLI quickstart. For Microsoft Copilot Studio, use Set up TextMine in Copilot Studio.

​
Adding a new MCP tool

Add the Public API V3 endpoint first. Then add a small tool class in the MCP service that implements the MCP tools marker interface, injects the Public API client, and exposes one or more tool methods. Keep provider-specific logic, OAuth flows, database access, validation, and business rules in the API service rather than the MCP service. The MCP layer should stay predictable for agents: tool name, tool description, input schema, API call, response.

​
Agent rules

  • Prefer named, narrow tools over general-purpose actions.
  • Read before mutating when the current state matters.
  • Ask for confirmation before creating, deleting, publishing, sending, assigning, or triggering workflow actions.
  • Use idempotency keys for retryable create or trigger operations when the API supports them.
  • Cite document, Vault, workflow, or run identifiers in agent output so humans can audit the action later.

​
When to use it

Use the TextMine MCP server when an external LLM client or automation needs a direct agent-friendly way to read or act on TextMine data. Use the web app when a human is operating interactively inside Workbench, Vault, Workflows, Playbooks, Records, Scribe, or Legislate. Start with For agents when choosing between these surfaces.
Last modified on July 7, 2026