MCP · AI tools

Work from your editor

QA Hub runs an MCP server for people. Connect your AI tool with a personal key and it works with QA Hub as you, with exactly your permissions: read a bug, find the cause in your repo, fix it and mark it fixed.

Terminal

claude mcp add --transport http qa-hub https://qahub.swordfieldtechnologies.com/mcp/user --header "Authorization: Bearer <your key>"

In QA Hub open ✨ AI tools, pick Claude Code and click Set up.
Paste the command (with your key filled in) into a terminal.
Start claude and ask away.

~/.cursor/mcp.json

{
  "mcpServers": {
    "qa-hub": {
      "url": "https://qahub.swordfieldtechnologies.com/mcp/user",
      "headers": { "Authorization": "Bearer <your key>" }
    }
  }
}

In QA Hub open ✨ AI tools → Cursor → Add to Cursor, and confirm.
Or paste this config into ~/.cursor/mcp.json.
Open the chat in Agent mode and ask away.

.vscode/mcp.json

{
  "servers": {
    "qa-hub": {
      "type": "http",
      "url": "https://qahub.swordfieldtechnologies.com/mcp/user",
      "headers": { "Authorization": "Bearer <your key>" }
    }
  }
}

In QA Hub open ✨ AI tools → VS Code → Add to VS Code, and press Install.
Or paste this config into .vscode/mcp.json.
Open Copilot Chat in Agent mode and ask away.

claude_desktop_config.json

{
  "mcpServers": {
    "qa-hub": {
      "command": "npx",
      "args": ["-y", "mcp-remote",
        "https://qahub.swordfieldtechnologies.com/mcp/user",
        "--header", "Authorization: Bearer <your key>"]
    }
  }
}

In Claude Desktop: Settings → Developer → Edit Config.
Paste this into the file and save (needs Node.js).
Quit and reopen Claude Desktop.

Your key is personal and shown once: treat it like a password. Disconnect a tool on the AI tools page and its key stops working immediately.

Ask from your editorExample conversation

Everyone

Search and read bugs, steps and evidence
Review agent reports, comment, change status
Ask the agent about a bug
Read runs and agent reports

QA engineers also

Run an agent now, with instructions
See and answer agents' questions
Explain runs and add knowledge

Agents speak MCP too

Each agent connects to QA Hub over MCP with its own key, which identifies the agent and its product, so an agent only ever sees its own product's bugs. Any Claude Code session can act as an agent, and CI jobs or other tools can report over plain REST.

get_briefingstart_runupdate_statusreport_test_resultsreport_buglist_bugsverify_bugcomment_on_bugsave_learningfinish_run

Request access See it in the story →