Why use MCP
Test management meets AI-assisted development
Create tests while you code
Implementing a new feature? Ask your AI assistant to create test cases as you build. No context switching, no copy-paste - tests are created directly in TestCollab.
Keep tests in sync with code
Refactored a function? Changed an API response? Your AI assistant can find and update affected test cases automatically, keeping documentation current.
Query coverage before changes
Before touching authentication code, ask what tests exist. Filter by priority, status, or keywords. Make informed decisions about test impact.
Open protocol, open source
Built on Anthropic's Model Context Protocol standard. Works with Claude, Cursor, Windsurf, Codex, and any MCP-compatible client. MIT licensed.
Understanding MCP
What is an MCP Server?
The Model Context Protocol (MCP) is an open standard created by Anthropic that lets AI assistants connect to external tools and data sources. An MCP server exposes capabilities that AI assistants can discover and use.
A bridge between AI and your tools
MCP servers expose tools, resources, and prompts that AI assistants can use. Think of it as giving your AI assistant hands to interact with external systems - databases, APIs, file systems, or in this case, test management.
The TestCollab MCP server exposes test case operations: list, create, update. Your AI assistant discovers these capabilities and uses them when relevant to your conversation.
Secure by design
MCP uses per-request authentication. Your API token is passed via headers - the server never stores credentials. Each session is isolated, and the AI assistant only accesses projects you authorize.
The protocol supports HTTP transport with standard auth patterns, making it enterprise-ready.
Compatible AI IDEs
Works with your favorite AI tools
Any MCP-compatible client can connect to the TestCollab MCP server. Here are the most popular ones.
Claude Code
Anthropic's AI coding assistant with native MCP support. Configure via claude_desktop_config.json and start managing tests from your terminal or desktop app.
Cursor
The AI-first code editor built on VS Code. Add TestCollab to your MCP servers and create tests while you write code - all in the same window.
Codex CLI
OpenAI's command-line coding agent with MCP support. Use --mcp-server-url to connect and manage tests directly from your terminal.
What you can do
MCP Server Capabilities
The TestCollab MCP server currently exposes three tools for test case management, with more planned.
Look up test cases
Search test cases with powerful filtering:
- Filter by title, priority, status, tags
- Sort by updated date, run count, failure rate
- Paginate through large result sets
- Get counts and metadata
Create new test cases
Create complete test cases:
- Title, description, priority
- Step-by-step instructions with expected results
- Tags and requirements linking
- Custom field support
Update existing tests
Update any test case field:
- Change title, steps, expected results
- Update priority or move to different suite
- Add or remove tags
- Partial updates supported
Setup guide
Getting Started
Connect in under 5 minutes. The MCP server supports both hosted and self-hosted deployment.
Option 1: Use hosted endpoint
Add TestCollab to your MCP client config:
Server URL: https://mcp.testcollab.io/mcp
Required headers:
- X-TC-API-Token: your-api-token
- X-TC-API-URL: https://api.testcollab.io
- X-TC-Default-Project: your-project-id
Get your API token from TestCollab → Profile → API Token. See the GitHub README for full config examples.
Option 2: Self-host
Clone and run the server locally for full control:
1. Clone the repo from GitHub
2. Run npm install and npm run build
3. Start with node dist/http-server.js
Server runs on port 3100 by default. See the README for Docker and environment configuration options.
FAQ
Answers teams look for
What is an MCP server?
An MCP (Model Context Protocol) server is a service that exposes tools, resources, and prompts to AI assistants. It's an open standard created by Anthropic that lets AI tools interact with external systems securely. The TestCollab MCP server lets AI assistants manage test cases.
Which AI assistants are supported?
Any MCP-compatible client works: Claude Code, Claude Desktop, Cursor, Windsurf, Codex CLI, and others. The protocol is open, so new clients are added regularly.
Is the MCP server open source?
Yes. The TestCollab MCP server is MIT licensed and available on GitHub at github.com/TCSoftInc/testcollab-mcp-server. You can self-host or use our hosted endpoint.
How does authentication work?
The MCP server uses header-based authentication. Pass your TestCollab API token via the X-TC-API-Token header. Tokens are scoped to your user permissions - the AI assistant can only access projects you have access to.
Can I use this with self-hosted TestCollab?
Yes. Set the X-TC-API-URL header to point to your self-hosted TestCollab instance. The MCP server proxies all requests to whatever API URL you configure.
What operations are supported?
Currently: list_test_cases (query with filters), create_test_case, and update_test_case. Suite management, test plans, and execution recording are on the roadmap.
Is there rate limiting?
The MCP server respects TestCollab API rate limits. For most use cases with AI assistants, you won't hit limits. Enterprise plans have higher quotas.
