# Traceable AI Connector (MCP)

> The official Model Context Protocol (MCP) connector for Traceable, from Traceable Docs. It lets Claude, or another AI assistant that accepts a custom server URL, read and author requirements documents, ID grids and trace links in your Traceable account, under your own permissions, with every change recorded against your name.

Canonical: https://traceabledocs.com//mcp

## Endpoint

- Server URL: https://app.traceabledocs.com/api/mcp/mcp
- Transport: streamable HTTP.
- Authorisation: OAuth 2.1. There is no API key to copy, paste or rotate.
- Your assistant discovers everything else on its own: where to send you to sign in, what permissions to ask for, and which tools exist.
- Nothing is readable before authentication.

## How to get access

### 1. An administrator turns it on

External agent access is off until an administrator of your organisation allows it, under Organisation settings, and then grants it to you individually under Members. Two controls, and both have to say yes.

### 2. You add the connector

In your assistant, add a custom connector pointing at the endpoint above. You will be sent to Traceable to sign in, if you are not already.

### 3. You approve the connection

Traceable shows a consent screen naming the application and exactly what it is asking for. Nothing is shared until you approve it, and you can revoke it later at any time.

## How to connect

- Claude (claude.ai, desktop, mobile): Settings, Connectors, Add custom connector, then paste https://app.traceabledocs.com/api/mcp/mcp
- Claude Code: `claude mcp add --transport http traceable https://app.traceabledocs.com/api/mcp/mcp`
- ChatGPT (developer mode) and Cursor: add the server URL as a custom connector or HTTP server.
- Clients that only speak stdio: `npx mcp-remote https://app.traceabledocs.com/api/mcp/mcp`

## What it can do

- Survey a project: its groups, documents, outlines and the trace graph between them.
- Read a document in full, or a slice of it, as Traceable Markdown.
- Create documents, optionally from one of your templates.
- Author and edit content: segments, blocks, ID grids, cells and columns.
- Create and delete trace links between TraceIDs, so coverage stays real.
- Rename a TraceID or a document prefix without breaking the links into it.
- Search and read your Reference Library, and cite an item on a document.
- Read your document Discussions and conversations, and post in them as you, marked as posted via the assistant.

## What it cannot do

- Reach anything you cannot reach yourself. It acts as you, under the same project and organisation membership as the web app.
- Edit a published document. Published versions are read-only to it, as they are to you.
- Publish, approve or sign anything. Review and eSignature stay with people.
- Delete a document that has not been archived first, mirroring the app.
- Act without a record. Every change it makes is audit-logged against your name and the application that made it.

## Tools

43 tools: 15 read only, 28 can change something. Claude groups them the same way: the read-only ones can be approved once, and the rest ask every time they are used.

### Capabilities

- `traceable_capabilities` (read): Get authoring capabilities

### Workspace

- `traceable_workspace_projects` (read): List organisations and projects
- `traceable_workspace_templates` (read): List templates
- `traceable_workspace_map` (read): Map a project
- `traceable_board_read` (read): Read the board

### Documents

- `traceable_doc_read` (read): Read document
- `traceable_doc_create` (write): Create document
- `traceable_doc_update_meta` (write): Update document metadata
- `traceable_board_placeholder_create` (write): Plan a placeholder card
- `traceable_board_placeholder_update` (write): Change a placeholder card
- `traceable_board_placeholder_delete` (write): Remove a placeholder card
- `traceable_doc_rename_prefix` (write): Rename document ID prefix
- `traceable_doc_archive` (write): Archive a document

### Content

- `traceable_segment_write` (write): Write segment
- `traceable_block_write` (write): Write block(s)
- `traceable_block_update` (write): Update block
- `traceable_block_delete` (write): Delete block
- `traceable_test_plan_read` (read): Read a Test Plan row
- `traceable_test_plan_entry` (write): Record Test Plan entry data
- `traceable_test_plan_evidence_add` (write): Add Test Plan evidence
- `traceable_test_plan_evidence_remove` (write): Remove Test Plan evidence
- `traceable_test_plan_evidence_upload` (write): Upload Test Plan evidence
- `traceable_image_upload` (write): Upload an image
- `traceable_image_view` (read): View an image

### Traceability

- `traceable_grid_column_add` (write): Add an ID-grid column
- `traceable_grid_column_rename` (write): Rename an ID-grid column
- `traceable_grid_column_set_type` (write): Change an ID-grid column type
- `traceable_grid_cell` (write): Set ID-grid cell
- `traceable_trace_id_rename` (write): Rename a TraceID
- `traceable_link_create` (write): Create a trace link
- `traceable_link_delete` (write): Delete a trace link

### Library

- `traceable_library_search` (read): Search or list the reference library
- `traceable_library_read` (read): Read reference library item
- `traceable_library_index` (write): Index reference library item
- `traceable_library_add_reference` (write): Add a reference
- `traceable_library_upload` (write): Upload a reference file
- `traceable_glossary_search` (read): Search the glossary
- `traceable_glossary_add_term` (write): Add a glossary term

### Skills

- `traceable_skill_list` (read): List skills
- `traceable_skill_read` (read): Read a skill

### Messages

- `traceable_thread_list` (read): List Discussions and conversations
- `traceable_thread_read` (read): Read a thread
- `traceable_thread_post` (write): Post in a thread

## Consent

The consent screen names the application asking, the Traceable account it would connect to, and the two permissions it wants: reading your projects and documents, and creating or editing document content. Nothing is shared until you approve.

Approving means document content leaves Traceable. The assistant receives the content it asks for, on your instruction, and handles it under your own agreement with whoever provides that assistant. You, or your organisation, are the party exporting that data. See section 6 of the Privacy Policy: https://app.traceabledocs.com/legal/privacy

## Turning it off

An administrator can switch external agents off for the whole organisation, which refuses every member at once, including anyone already connected. They can also revoke one person, or one connection, from the Members panel. Both take effect immediately.

## Plans

The connector is included on every plan, with rate limits on Free. Prices: https://traceabledocs.com//pricing.md

## Links

- Working with an AI assistant (MCP): https://traceabledocs.com//docs/ai-assistant-mcp
- MCP connector reference: https://traceabledocs.com//docs/mcp-reference
- Why MCP matters for requirements work: https://traceabledocs.com//requirements-management-software-with-mcp
- Contact: https://traceabledocs.com//contact
