Ask your test data: embedlabs-mcp preview

The questions engineers ask about test and requirement data are plain: which requirements have no test, how did the benches do this month, what failed in last night’s run. Today the answers sit behind dashboards and filters. embedlabs-mcp lets an AI assistant ask Bud and Bloom directly, through the Model Context Protocol.

Two answers from the demo project

Asked “Which PRJ requirements have no verifying test case?”, the client called bloom_traceability_coverage_gaps and answered: 1 of 9 requirements is uncovered, PRJ-REQ-009 “Checkout confirmation timeout” (High); project coverage is 88.9 %.

Asked “How did our test runs go in the last 30 days?”, it called bud_runs_stats: 12 runs, 7 passed and 5 failed (58.3 %); across 52 tests, 86.5 % passed.

Both answers came from the applications’ own APIs, with the asking user’s permissions.

What is inside

• 86 tools: 64 for Bloom, 22 for Bud, each marked read-only or data-changing.
• Browser sign-in with a request code and PKCE; tokens kept in the system credential store and renewed automatically.
• The same tools as bud and bloom terminal commands with JSON output.
• No database access and no telemetry. Requests go only to your instances, and every call shows up in their audit log with an embedlabs-mcp user agent.

Try it early

The server needs Bud or Bloom 2.0.0 or later and is not released yet. The documentation, including a reference of every tool, is on this site. If you want early access, tell us which questions you would ask first.