boss-baby by anix-lynch - MCP Server

👶💼 Project: Boss Baby

"I just want my AI workforce to remember me."

🧠 Tagline

Long-term memory for your AI dev stack — no cache, no cloud, no VM drama.

🚀 Overview

Boss Baby is a personal AI Memory Sandbox that teaches local and IDE-based AIs to remember you permanently.

✅ Phase 1 (Complete) — Symbolic Memory
🚧 Phase 2 (In Construction) — Semantic Memory "Mini Animal LLM" 🐣

It runs entirely on localhost and acts as:

🧩 API & MCP bridge between AIs
🧠 Shared memory layer for Cursor, Claude Code, Windsurf, Continue
🧰 Testing ground for agent context recall and AI personality persistence

🧰 Tool Stack (The AI Workforce)

Category	Tools
🧑‍💻 AI Coders	Claude Code, ChatGPT, Kilocode, Cline, Roo Code, Aider CLI
🧠 AI IDEs	Cursor AI, Windsurf, Continue.dev, Zed
⚡ Terminals & Runtimes	Warp AI Terminal, LM Studio
🐙 Infra & Versioning	GitHub Codespaces
👩‍💼 Human Boss	Bchan — the OG memory architect 💅

🧩 Architecture

resume.yaml (Source of Truth)
├──→ api/main.py        → FastAPI endpoint http://127.0.0.1:8000/resume
└──→ mcp/server.py      → MCP memory tools for IDEs

Both share the same data source (resume.yaml) for unified, local-only context brain.

🧠 Features

🔐 100% local (127.0.0.1)
🧩 FastAPI endpoint + MCP server
⚙️ CLI scripts with sanity checks
🤖 IDE-ready memory config (mcp.json)
💾 No cloud or VM dependency
🧱 Phase 2 incoming: semantic embedding layer for natural phrase understanding

📁 Project Structure

boss-baby/
├── README.md                    (this file)
├── START_HERE.md               (quick start)
├── ai-config/
│   ├── base/data/
│   │   └── resume.yaml         (your memory source)
│   └── continue/
│       ├── config.json         (Continue.dev config)
│       ├── main.py             (FastAPI server)
│       ├── mcp_server_v2.py    (MCP server)
│       └── Makefile            (all commands)
└── scripts/
    ├── run_api.sh              (start FastAPI)
    ├── run_mcp.sh              (start MCP)
    └── test.sh                 (run tests)

🧰 Quick Commands (Warp-ready)

# Start API
cd ai-config/continue
make run              # Start FastAPI at http://127.0.0.1:8000/resume

# Test API
make test-api         # GET /resume endpoint

# Test MCP
make test-mcp         # Run MCP server tests

# See all commands
make help

🧠 Example Use Cases

Tool	Command
Claude Code	`@resume` mention or `/resume` command
Cursor AI	Type `/resume` in chat
Continue.dev	`@resume.yaml` context provider
Warp Terminal	`curl http://127.0.0.1:8000/resume`
MCP Clients	Use any of 7 available tools

🔮 Phase 2 — Semantic Memory (Coming Soon)

Teach AIs to recognize you, not just your files.

🐣 Mini LLM Trainer — uses sbert MiniLM for local embeddings
🧬 aliases.yaml + training.json → vector store embeddings.faiss
⚙️ MCP integration → semantic intent resolver ("get Boss Baby resume" → get_resume_full)
🎯 Goal: Build a local semantic fingerprint for Bchan across all IDEs

🏆 Why It Matters

Goal	Outcome
🧠 AI context sync	Every IDE remembers the same brain
🧩 Cross-AI testing	Benchmark coder agents on context retention
💡 API practice	Hands-on FastAPI + MCP architecture
🔒 Ownership	No cloud memory lock-in — you own your context

👩‍💼 Your Memory Data

Edit this file to update your memory everywhere:

ai-config/base/data/resume.yaml

Contains:

📋 Certifications
💼 Experience
🚀 Projects
📞 Contact Info

👩‍💼 Credits

Creator: Anix "Bchan" Lynch
Studio: GoZeroShot.dev
Inspiration: Every AI that ever forgot me.

🍼 Boss Baby

Because every AI team needs a boss with permanent memory.

🚀 Get Started

Read: START_HERE.md
Run: make test-api (verify HTTP API)
Run: make test-mcp (verify MCP server)
Use: In your favorite AI IDE → @resume or /resume

Everything is local. Everything is yours. Everything remembers. 🧠