Skip to content
EmbedLabs
GitHub
documentation

Introduction

What embedlabs-mcp is, how it connects an MCP client to Bud and Bloom, and what it never does.

embedlabs-mcp is a local program that connects an MCP client, such as Claude Desktop, Claude Code or Codex, to your self-hosted Bud TMP and Bloom PLM instances. The client’s model calls its tools; the tools call the Bud and Bloom APIs as you.

MCP client (Claude, Codex, ...)
|
v
embedlabs-mcp (runs on your machine)
| |
v v
Bloom API Bud API (your instances)
  • Bloom tools (bloom_*, 64): projects, requirements, test cases, test suites, campaigns, typed links, the traceability matrix, impact analysis, coverage gaps, defects, designs, risks, baselines, search, controlled documents and their review workflow.
  • Bud tools (bud_*, 22): products, test runs, statistics, custom runs, the test catalog, the execution queue, Test Station status, results and artifacts, and publishing a run to Bloom.
  • Sign-in tools (6): browser sign-in for each product.
  • Terminal commands: bud and bloom, the same tools without an MCP client.
  • It never touches either PostgreSQL database. Every operation goes through the application APIs, so Bud’s and Bloom’s permissions, business rules and audit log stay in the loop.
  • It sends no telemetry and talks to no EmbedLabs service. Requests go only to the URLs you configure.
  • It does not record reviews or approvals. Signing off a controlled document stays in the Bloom web app.
  • Node.js 22.23.4 or a later 22.x, 24.22.1 or a later 24.x, or 26.12.1 or later
  • A Bud or Bloom instance and an account on it

Next: install the server.