Engineering MCP

Local MCP for engineering: tickets, QA evidence, and time go in; investigation closes root cause; knowledge stays; reporting comes out ready.

It doesn't replace what you already have:

ENGINEERING MCP
      │
      ├── WORK        tickets · tasks · projects
      ├── QA          testing evidence · bugs · artifacts
      └── TIME        tracking · estimates · metrics
              │
              ▼
        INVESTIGATION
              │
              ▼
          KNOWLEDGE
              │
              ▼
          REPORTING

Problem it solves

A Jira ticket, a CI failure, and a time entry live in three places. When status time comes, no one puts the package together.

This MCP records the cycle in the same local database:

1. Opens the ticket (WORK)

2. Attaches the fail / bug / screenshot (QA) — without running the suite

3. Logs hours and estimate (TIME)

4. Investigates to root cause (INVESTIGATION)

5. Becomes a playbook (KNOWLEDGE)

6. report_ticket returns the package for handoff, daily, or career evidence

Related MCP server: Productivity Tracker MCP Server

Requirements

• Node 22+ (native node:sqlite, same as qa-oracle)

• MCP client (Cursor)

Installation (no absolute path)

Cursor does not point to the Desktop folder. It downloads the package — same as qa-lab-agent. The database lives in ~/.engineering-mcp/engineering.db, on any machine.

Config in Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "engineering": {
      "command": "npx",
      "args": ["-y", "engineering-mcp"]
    }
  }
}

No args with /Users/.... No ENGINEERING_MCP_DB. Reload the MCPs.

This works after the package is on npm or GitHub (below). Until publishing, use the local shortcut (also without a path in mcp.json):

cd ~/Desktop/engineering-mcp
npm install
npm install -g .

{
  "mcpServers": {
    "engineering": {
      "command": "engineering-mcp"
    }
  }
}

npm install -g . puts the command on the PATH. The mcp.json only mentions the name.

Where to host (npm vs GitHub vs cloud)

No need for your own site. The “site” is npmjs.com — the same as mcp-lab-agent. Cursor runs the MCP on your machine; the cloud only delivers the code.

Cloud like URL ("url": "https://...") exists in Cursor, but then SQLite is no longer local: tickets, bugs, and hours would live on a server. That's a multi-user product, not this v0.

Publishing to npm (same as lab-agent)

The name engineering-mcp is free. With an npm account already logged in:

cd ~/Desktop/engineering-mcp
npm test
npm publish --access public

After that, any Mac just uses the npx block above.

Publishing to GitHub (without npm)

cd ~/Desktop/engineering-mcp
git init
git add .
git commit -m "feat: engineering mcp v0.1"
gh repo create Wesley-Gomes93/engineering-mcp --public --source . --remote origin --push

Alternative config, still without a path:

{
  "mcpServers": {
    "engineering": {
      "command": "npx",
      "args": ["-y", "github:Wesley-Gomes93/engineering-mcp"]
    }
  }
}

Restart Cursor (or reload the MCPs). The database is created on its own in ~/.engineering-mcp/.

Typical loop

"Abre um projeto Atlas (ATL) e um bug P1: timeout no checkout"
  → work_upsert_project + work_upsert_ticket

"Registra o fail do checkout.spec.js e anexa o log"
  → qa_record_run + qa_attach_evidence

"Estima 4h e lança 1.5h de repro"
  → time_estimate + time_log

"Abre investigação: hipótese de seletor instável"
  → investigate_open + investigate_add_finding

"Fecha como flaky: timing no botão Finalizar"
  → investigate_conclude

"Vira playbook: wait da animação + data-testid"
  → knowledge_save

"Me dá o pacote do ticket"
  → report_ticket

If you don't know the tool: eng_route with the phrase in Portuguese.

Tools

WORK

Status: backlog · todo · doing · review · done
Priority: p0 … p3
external_key stores the Jira/GitLab key without pulling an API yet.

QA

Running tests still happens in qa-lab-agent. Corporate Jira history stays in qa-oracle.

TIME

INVESTIGATION

Classification: bug · flaky · infra · regression · unknown

KNOWLEDGE

REPORTING

Structure

engineering-mcp/
├── src/
│   ├── server.js              # MCP stdio
│   ├── lib/store.js           # SQLite — fonte da verdade
│   └── domains/
│       ├── work.js
│       ├── qa.js
│       ├── time.js
│       ├── investigation.js
│       ├── knowledge.js
│       └── reporting.js
└── test/store.test.js

Default database: ~/.engineering-mcp/engineering.db (doesn't travel with the repo).

Nothing leaves the machine beyond what the agent already sees in the chat. No Jira/GitLab in v0 — external_key is the hook for sync later.

v0 vs later

Now: local database, closed loop, Cursor.

Later (only if v0 is used):

• Jira → ticket sync (reuse qa-oracle client)

• Import run from qa-lab-agent / GitLab job

• HTML dashboard from report_status

Security

• Database in ~/.engineering-mcp/ (outside the repo)

• No tokens in v0

• Publishing to npm/GitHub sends code, not tickets