# Engineering Os > Your AI finally remembers your codebase - **Type:** MCP server - **Install:** `agentstack add mcp-engineering-os-engineering-os` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [engineering-os](https://agentstack.voostack.com/s/engineering-os) - **Installs:** 0 - **Category:** [Integrations](https://agentstack.voostack.com/c/integrations) - **Latest version:** 0.1.0 - **License:** Apache-2.0 - **Upstream author:** [engineering-os](https://github.com/engineering-os) - **Source:** https://github.com/engineering-os/engineering-os - **Website:** https://engineering-os.github.io/engineering-os/ ## Install ```sh agentstack add mcp-engineering-os-engineering-os ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # Engineering OS [](https://github.com/engineering-os/engineering-os) [](https://www.npmjs.com/package/engineering-os) [](./LICENSE) **Your AI finally remembers your codebase.** Stop re-explaining your architecture every session. Engineering OS gives Claude Code, Cursor, Codex, and any MCP-compatible AI code editor the memory they've always been missing. They finally write code like someone who actually knows your project. ```bash npm install -g engineering-os cd your-project eos init --claude ``` That's it. Your AI now has persistent project knowledge. Permanently. --- ## The Problem Every new session, your AI re-explores files, re-debates decisions you already made, and has no idea your frontend talks to three backend services. | Today (without Engineering OS) | With Engineering OS | | --------------------------------------------- | ------------------------------------------------ | | "Let me explore your project structure..." | Already knows. Starts working immediately. | | "Should we use Redux or Zustand?" | "You decided on Zustand last month. Here's why." | | Suggests code that breaks another service | Sees the dependency graph. Avoids breakage. | | You explain the same gotcha every session | Learned it once. Warns you proactively. | | grep-ing for 30s trying to find that function | Instant answer. It's pre-indexed. | | Security review is something you do "later" | Already scanned. Flags issues as you code. | ## How It Works ``` AI Code Editor (Claude Code / Cursor / Codex / Any MCP Client) │ MCP Protocol (stdio) ▼ Engineering OS (runs on your machine, zero cloud) ├── Code Search (FTS5 + BM25, instant results) ├── Service Graph (cross-repo topology) ├── Skill Memory (learns from every session) ├── Decision Records (never re-debates) ├── Convention Enforcement (your rules, always followed) └── Security Scanner (OWASP, CVE, compliance) │ ▼ Your Codebase + Linked Repos (SQLite, 100% local) ``` 1. `eos init` indexes your codebase, discovers architecture, builds the service graph 2. Your AI tool connects automatically via MCP (no manual config) 3. AI calls `eos_context` and gets full project knowledge instantly 4. AI calls `eos_learn` when it discovers something. That knowledge persists forever. ## Quick Start ```bash # Install npm install -g engineering-os # Point it at your project cd your-project eos init --claude --cursor --codex # Link sibling services (optional) eos link backend-api ../backend-api eos link ml-pipeline ../ml-service # Keep it fresh eos refresh --incremental ``` ## What You Get - **Never explains twice.** Architecture, routes, conventions, and decisions persist across every session. - **Sees all your repos.** Link your mobile app, BFF, backend, ML pipeline. Your AI sees the full picture. - **Learns your tricks.** That weird workaround? That gotcha with uploads? Remembered. Forever. - **Catches security issues.** Secrets, injection flaws, known CVEs in your dependencies. Scans continuously. - **Finds code instantly.** Pre-indexed search. Ask "where's the auth middleware?" and get the answer in milliseconds. - **Maps your architecture.** Auto-detects how your services connect, which layers exist, what patterns you follow. ## CLI Commands | Command | What it does | | ---------------------------- | ---------------------------------------------- | | `eos init --claude --cursor --codex` | Index your project + generate AI context files | | `eos serve` | Start the MCP server | | `eos refresh --incremental` | Update knowledge after code changes | | `eos cursor generate` | Regenerate Cursor rules and EOS skills | | `eos codex generate` | Regenerate Codex `AGENTS.md`, skills, and MCP config | | `eos link ` | Link a sibling repository | | `eos workspace init` | Create team-shared config (checked into git) | | `eos index --watch` | Continuous re-indexing as you code | | `eos status` | Show knowledge health and drift | ## Works With - **Claude Code** (Anthropic) - **Cursor** - **Codex** (OpenAI) - **VS Code** - Any editor that speaks MCP protocol ## Team Sharing Create `eos.workspace.yaml` and check it into git. Your entire team shares the same conventions and decisions: ```yaml name: my-project type: monorepo org: my-company repos: - name: api-gateway path: ../api-gateway role: gateway - name: backend path: ../backend-service role: bff conventions: - name: error-handling rule: "Use Result pattern. Only controllers throw HTTP exceptions." - name: state rule: "Zustand only. No Redux. One store per domain." decisions: - title: "Redis for sessions" decision: "Redis Cluster for session management" rationale: "Sub-ms reads, horizontal scaling, built-in expiry" ``` ## Troubleshooting ### MCP Error `-32000` in Claude Code Node.js version mismatch. The native SQLite module was compiled for a different Node version. ```bash # Fix: rebuild for your current Node npm rebuild better-sqlite3 # Or just reinstall npm install -g engineering-os ``` **Why:** If you use nvm and your shell defaults to a different Node version than what installed Engineering OS, the native binary won't load. Make sure both match. ### Requirements - Node.js >= 18 (LTS 20 or 22 recommended) - No API keys, no cloud, nothing external ## Documentation Full docs with API reference, guides, and tool list: [engineering-os.github.io/engineering-os](https://engineering-os.github.io/engineering-os/) ## Architecture ``` packages/ ├── shared/ # Shared types ├── core/ # MCP server, knowledge engine, security, workflows ├── cli/ # CLI (the `eos` command) ├── adapter-claude/ # Claude Code plugin (11 skills) ├── adapter-cursor/ # Cursor rules generator ├── adapter-vscode/ # VS Code extension └── engineering-os/ # Single npm package (bundles everything) ``` ## License Apache 2.0 — see [LICENSE](./LICENSE) ## Source & license This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [engineering-os](https://github.com/engineering-os) - **Source:** [engineering-os/engineering-os](https://github.com/engineering-os/engineering-os) - **License:** Apache-2.0 - **Homepage:** https://engineering-os.github.io/engineering-os/ Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/mcp-engineering-os-engineering-os - Seller: https://agentstack.voostack.com/s/engineering-os - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.