Connecting Claude to Your TMS via MCP
Connecting Claude to your TMS via the Model Context Protocol (MCP) lets an AI agent read and manage your test artifacts directly — creating test cases, organizing suites, updating runs, and reading results — instead of you doing each action by hand in the UI. QAM Hub ships an MCP server that exposes its core operations as tools an agent can call.
This guide explains what MCP is, what becomes possible when Claude can drive a test management system, and how to get access to the QAM Hub MCP server.
What is the Model Context Protocol?
The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools and data sources through a consistent interface. Instead of an AI model being limited to text in a chat window, an MCP server exposes a set of tools — concrete actions the model can call — so the assistant can actually do things in another system.
Applied to test management, this means an agent like Claude is no longer just advising you about testing in the abstract. Through the QAM Hub MCP server, it can operate on the real test artifacts in your project.
What Claude can do in QAM Hub through MCP
The QAM Hub MCP server exposes its core test-management operations as agent-callable tools. Through it, Claude can work with:
- Test cases — create cases one at a time or in batches, update them, fetch them, manage tags, reorder fields, and list or restore versions.
- Suites and groups — create, update, delete, and list the structure your cases live in.
- Checklists — create and update checklist items and runs, including CSV import and export.
- Test runs — create and update runs and set execution statuses, including in bulk.
- Requirements — create, update, and list requirements, with CSV import and export.
- Milestones — create, update, and list milestones.
- Custom fields, tags, and templates — create and manage the metadata that structures your test base.
In other words, most of what a person does in the QAM Hub interface, an agent can do through MCP — at the pace and scale of automation.
What this makes possible
Connecting an AI agent to your TMS changes the kind of work you can hand off. A few practical examples:
- Generate a test suite from a spec. Point Claude at a requirements document or a feature description and have it draft a structured set of test cases directly into a suite — for a QA engineer to review, not retype.
- Bulk-organize an existing test base. Ask the agent to retag, regroup, or restructure cases across a project, instead of editing them one by one.
- Turn results into action. Have the agent read a run's results and create requirements, follow-up cases, or a summary from what failed.
- Migrate and import. Use the agent to shape and import cases or requirements from CSV into the right structure.
The throughput of automation, applied to test management work that used to be entirely manual.
Keeping a human in the loop
Letting an agent write to your test management system only works if a person stays in control. Two things make that safe in QAM Hub:
- Version history. Every change to a test case is snapshotted, so anything an agent does can be reviewed and rolled back. See version history and rollback.
- Scoped access. The agent operates with an API token tied to a specific project and role, so its reach is limited to what you grant — see roles and permissions.
This reflects how QA Madness approaches AI in testing generally: the agent proposes and executes the repetitive work, and a QA engineer makes the decisions. We cover that philosophy in AI test management tools.
How MCP fits with the rest of QAM Hub
MCP is one of several ways QAM Hub connects to automation and AI. Automated results flow in from Playwright and Cypress reporters, AI features generate and analyze cases inside the app, and agent flows run autonomous test executions. MCP is the layer that lets an external AI agent orchestrate across all of it programmatically.
Getting access to the QAM Hub MCP server
The QAM Hub MCP server is currently available to QAM Hub users on request, rather than as a public download, so we can support each setup directly while the integration matures. If you would like to connect Claude (or another MCP-compatible agent) to your QAM Hub project, contact the QA Madness team to request access and setup guidance.
Once connected, you generate an API token in QAM Hub (Profile → API tokens), point the MCP server at your project, and your agent can begin working with your test artifacts.
Frequently asked questions
What is MCP in the context of test management?
MCP (Model Context Protocol) is an open standard that lets an AI agent call tools in an external system. For test management, it means an agent like Claude can create test cases, manage suites and runs, and read results directly in the TMS, rather than only discussing testing in chat.
What can Claude actually change in QAM Hub?
Through the MCP server, Claude can create and update test cases, suites, runs, requirements, milestones, checklists, custom fields, tags, and templates — most of the operations available in the QAM Hub interface, including batch actions.
Is it safe to let an AI agent write to my TMS?
Yes, with controls. The agent uses an API token scoped to a specific project and role, and every test-case change is versioned and reversible, so a human can review and roll back anything the agent does.
How do I get the QAM Hub MCP server?
It is available to QAM Hub users on request while the integration matures. Contact the QA Madness team to request access and setup guidance for connecting Claude or another MCP-compatible agent.
Summary
The Model Context Protocol turns a test management system from something you operate by hand into something an AI agent can operate with you. Connected to QAM Hub, Claude can create cases, organize suites, manage runs, and read results programmatically — with version history and scoped access keeping a human firmly in control.
QAM Hub is a test management system built by QA Madness, a software testing and QA automation company. Its MCP server lets AI agents like Claude manage test artifacts directly — part of a human-in-the-loop approach to AI-assisted QA, built by engineers who do this work every day.