Testomat.io MCP server

Testomat.io MCP Server

Model Context Protocol (MCP) server that enables AI assistants (Claude, Cursor, OpenCode, etc.) to interact with Testomat.io Public API v2.

Features

• One CLI-style tool per entity - tests, suites, runs, testruns, plans, rungroups, steps, snippets, labels, requirements, branches, tags, milestones, issues; each takes a command argument (list, get, create, update, delete, plus scoped issues_*/attachments_*/share commands where applicable) and flat command params
• Project Information - fetch project configuration, metadata, features, and CI profiles
• Issue Linking - link/unlink issues to any resource (issues_* commands on each entity, plus the global issues tool)
• Attachments - attachments_* commands on tests, suites, and testruns; upload sends a local file path as multipart field files
• Requirements - including file uploads from local file paths
• API Compatibility - automatic handling of payload format differences (flat vs wrapped)
• Automatic API Sessions - groups MCP changes in Testomat.io history using API sessions
• Run Management - status transitions via status_event parameter
• TQL-Only Search - tests/runs with command: "list" use tql as the single search/filter input
• Built-In TQL Reference - TQL parameters include the exact field whitelist and examples; tql_help provides syntax details on demand
• Tool Surface Profiles - expose only the tools a session needs via --tools full|core|read (default full); cuts the per-call schema cost for long agentic sessions

Quick Start

Installation

npm install -g @testomatio/mcp@latest

Need enterprise analytics tools? Install @testomatio/mcp-enterprise@latest instead. Details are in the Enterprise Analytics section below.

Configuration

Required credentials:

• TESTOMATIO_PROJECT_TOKEN - Your project API token
• TESTOMATIO_PROJECT_ID - Your project ID

Run server:

npx testomatio-mcp --token <PROJECT_TOKEN> --project <PROJECT_ID>

Or with environment variables:

export TESTOMATIO_PROJECT_TOKEN=<PROJECT_TOKEN>
export TESTOMATIO_PROJECT_ID=<PROJECT_ID>
npx testomatio-mcp

A bare hostname is expanded to https://<host>. For full control use --base-url / TESTOMATIO_BASE_URL, which takes precedence over the host option

Optional: tool surface profile

By default the server exposes all tools. For long, token-sensitive sessions you can expose a smaller set with --tools:

npx testomatio-mcp --token <PROJECT_TOKEN> --project <PROJECT_ID> --tools core

Profile	What's exposed
full (default)	Everything
core	Core entities with all commands (excludes the steps, snippets, labels, rungroups tools)
read	Core entities restricted to read-only commands (list, get, search, issues_list, attachments_list)

Values are case-insensitive; an unknown value prevents the server from starting. Set the profile at launch with the flag or the TESTOMATIO_TOOLS environment variable — it can't be changed mid-session. The CLI flag takes precedence when both are set.

Usage with AI Assistants

Cursor IDE

Add this config to .cursor/mcp.json in your project, or to ~/.cursor/mcp.json for global access:

{
  "mcpServers": {
    "testomatio": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@testomatio/mcp@latest",
        "--token",
        "<TOKEN>",
        "--project",
        "<PROJECT_ID>"
      ],
      "env": {
        "TESTOMATIO_BASE_URL": "https://app.testomat.io"
      }
    }
  }
}

Claude Desktop

Add this config to:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "testomatio": {
      "command": "npx",
      "args": [
        "-y",
        "@testomatio/mcp@latest",
        "--token",
        "<TOKEN>",
        "--project",
        "<PROJECT_ID>"
      ],
      "env": {
        "TESTOMATIO_BASE_URL": "https://app.testomat.io"
      }
    }
  }
}

OpenCode

Add this config to opencode.json in your project root, or to ~/.config/opencode/opencode.json for global access:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "testomat": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "@testomatio/mcp@latest",
        "--token",
        "<TOKEN>",
        "--project",
        "<PROJECT_ID>"
      ],
      "enabled": true,
      "environment": {
        "TESTOMATIO_BASE_URL": "https://app.testomat.io"
      }
    }
  }
}

HTTP Transport

Besides stdio, the server runs over Streamable HTTP on a Cloudflare Worker hosted by Testomat.io. The project is part of the URL, so every tool signature stays the same:

https://mcp.testomat.io/mcp/<project_id>

Point an MCP client at that URL with a project token:

{
  "mcpServers": {
    "testomatio": {
      "url": "https://mcp.testomat.io/mcp/<PROJECT_ID>",
      "headers": {
        "Authorization": "Bearer <PROJECT_TOKEN>"
      }
    }
  }
}

Web connectors such as claude.ai have nowhere to put a static token and instead run OAuth 2.1 with PKCE and Dynamic Client Registration against the same URL. Tokens starting with testomat_ or tstmt_ always bypass OAuth and are passed straight through, so IDE clients and CI keep working.

The endpoint is POST only; GET returns 405, because the server never initiates traffic. Testing with curl requires both media types in Accept:

curl -sS https://mcp.testomat.io/mcp/<PROJECT_ID> \
  -H "Authorization: Bearer <PROJECT_TOKEN>" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Self-hosted installations keep using stdio.

Quick Examples

Every entity tool works like a CLI: pass command plus the params that command needs. Each param description in the tool schema lists the commands it applies to.

List tests:

{
  "name": "tests",
  "arguments": { "command": "list", "page": 1, "per_page": 50, "tql": "priority == 'high'" }
}

Create test:

{
  "name": "tests",
  "arguments": {
    "command": "create",
    "title": "User login test",
    "suite_id": "123",
    "priority": "high"
  }
}

Create run:

{
  "name": "runs",
  "arguments": {
    "command": "create",
    "title": "Smoke tests",
    "kind": "automated",
    "env": "production"
  }
}

Finish run:

{
  "name": "runs",
  "arguments": {
    "command": "update",
    "run_id": "456",
    "status_event": "finish"
  }
}

Upload attachment to a test:

{
  "name": "tests",
  "arguments": {
    "command": "attachments_upload",
    "test_id": "123",
    "file_path": "/path/to/screenshot.png"
  }
}

Bulk create/update tests from markdown:

{
  "name": "tests",
  "arguments": {
    "command": "bulk_upsert",
    "markdown": "<!-- suite\nid: @S380c64db\n-->\n# Login Functionality\n<!-- test\nid: @T12345678\npriority: high\n-->\n# Successful Login\n<!-- test -->\n# Failed Login"
  }
}

Tests with an id: @T... are updated, tests without one are created; suites are resolved by id: @S... or title and created when missing. Up to 100 tests per call. See docs/tools.md for the full semantics.

Documentation

Complete tool reference: docs/tools.md

Project Structure

src/
|- config/          # Config loading, defaults
|- core/            # Errors, logger
|- api/             # HTTP client, Testomat.io API client
|- mcp/             # MCP server, tools, registry
|  |- definitions/  # Tool definitions by entity
|  |- configs/      # Registry generation configs
|  `- registry/     # Tool handlers
`- cli/             # CLI bootstrap

Environment Variables

Variable	Required	Default	Description
TESTOMATIO_PROJECT_TOKEN	Yes*	-	Project token (preferred)
TESTOMATIO_API_TOKEN	Yes*	-	Alternative token
TESTOMATIO_PROJECT_ID	Yes	-	Project ID
TESTOMATIO_HOST	No	-	API host, e.g. beta.testomat.io
TESTOMATIO_BASE_URL	No	https://app.testomat.io	API base URL, wins over TESTOMATIO_HOST
TESTOMATIO_TOOLS	No	full	Tool profile: full, core, or read

*Either TESTOMATIO_PROJECT_TOKEN or TESTOMATIO_API_TOKEN

Corporate TLS Certificates

If the MCP server runs behind a corporate proxy or TLS inspection, Node.js may reject Testomat.io HTTPS requests even when the same URL works in a browser. This usually means the company root certificate is trusted by the operating system, but not by Node.js.

Use Node.js system CA support:

NODE_OPTIONS=--use-system-ca testomatio-mcp --token <TOKEN> --project <PROJECT_ID>

For MCP clients, pass NODE_OPTIONS in the server environment:

{
  "mcpServers": {
    "testomatio": {
      "command": "testomatio-mcp",
      "args": ["--token", "<TOKEN>", "--project", "<PROJECT_ID>"],
      "env": {
        "NODE_OPTIONS": "--use-system-ca"
      }
    }
  }
}

If your Node.js version does not support --use-system-ca, export the corporate root certificate to a PEM file and use Node.js extra CA support:

NODE_EXTRA_CA_CERTS=/path/to/company-root-ca.pem testomatio-mcp --token <TOKEN> --project <PROJECT_ID>

Important Notes

• CLI-style Tools - One tool per entity; all operations are the command argument (tests + command: "list" instead of the old tests_list). Unknown commands return an error listing the valid ones
• Run Status - Use runs with command: "update" and status_event for transitions (finish, launch, rerun, etc.)
• Search/Filter - No dedicated /search endpoints; filtering is done via the list command (tql for tests and runs, OpenAPI-aligned filters for other entities)
• Slim List Responses - List commands request compact API responses by default and omit heavy entity fields and null values. Pass verbose: true to return full objects, or fields: ["id", "title", "description"] to return only selected fields. Both options disable the backend slim=true request so heavy fields remain available when requested.
• TQL - Use tql as the single search/filter input for tests/runs with command: "list"
• TQL Syntax - For user-facing syntax details and more examples, see the official TQL docs: https://docs.testomat.io/advanced/tql/
• TQL Scope - TQL parameter descriptions keep the documented field whitelist in-band; call tql_help for syntax details and additional examples
• Issue Linking - Scoped commands available on each entity: command: "issues_link" / "issues_unlink"
• Attachments - Scoped commands available on tests, suites, and testruns: command: "attachments_list" | "attachments_upload" | "attachments_delete". Upload sends one local file path as multipart field files.
• Enterprise Package - Analytics tools are intentionally exposed only by @testomatio/mcp-enterprise, not by the standard @testomatio/mcp package
• API Sessions - The server automatically starts a Testomat.io session before the first POST, PUT, or DELETE request, sends the returned session hash as X-Session-Hash on later mutating requests, and stops the session when the MCP server shuts down. GET requests do not start or use sessions.

Development

npm install
npm run start -- --token <TOKEN> --project <PROJECT_ID>
npm test

For local MCP development, point Claude Desktop to the checked-out entrypoint:

{
  "mcpServers": {
    "testomatio-local": {
      "command": "node",
      "args": ["/path/to/mcp/index.js", "--token", "<TOKEN>", "--project", "<PROJECT_ID>"]
    }
  }
}

Enterprise Analytics

Enterprise analytics is available only in the separate @testomatio/mcp-enterprise package.

Installation:

npm install -g @testomatio/mcp-enterprise@latest

Run:

testomatio-mcp-enterprise --token <PROJECT_TOKEN> --project <PROJECT_ID>

Included tools:

• analytics_tests - GET /api/v2/{project_id}/analytics/tests/{kind}
• analytics_stats - GET /api/v2/{project_id}/analytics/stats/{kind}
• analytics_charts_list - GET /api/v2/{project_id}/analytics/charts
• analytics_charts_get - GET /api/v2/{project_id}/analytics/charts/{id}
• analytics_charts_results - GET /api/v2/{project_id}/analytics/charts/{id}/result

Analytics endpoints require the api_analytics subscription feature. Use q as the TQL filter parameter for analytics tools.

Example analytics_tests call:

{
  "name": "analytics_tests",
  "arguments": {
    "kind": "flaky",
    "q": "priority == 'high'",
    "days": 30,
    "page": 1,
    "per_page": 20
  }
}

Example analytics_stats call:

{
  "name": "analytics_stats",
  "arguments": {
    "kind": "success-rate-by-date",
    "q": "tag IN ['@smoke']",
    "from": "2026-04-01",
    "to": "2026-04-30"
  }
}

analytics_charts_get returns a chart definition with its TQL queries; analytics_charts_results returns per-query totals or the matching tests/runs of one query (by its zero-based number).

Example analytics_charts_results call:

{
  "name": "analytics_charts_results",
  "arguments": {
    "chart_id": "abc123",
    "number": 2,
    "per_page": 50
  }
}