How to give Claude Code persistent memory with 51 MCP tools

#mcp #ai #claude #tutorial

The Problem

Claude Code uses MEMORY.md files for persistence. 200-line limit, no search, no validation, no multi-agent support. After months of fighting this, I built BrainDB.

Setup (2 minutes)

Clone, configure, start:

git clone https://github.com/beckfexx/BrainDB.git
cd BrainDB && cp .env.example .env
bun install && bun run start

Add to your MCP config (Claude Code settings):

{

  "mcpServers": {

    "braindb": {

      "command": "bun",

      "args": ["run", "src/mcp-client.ts"],

      "env": { "BRAINDB_URL": "http://localhost:3197" }

    }

  }

}

What you get (51 tools)

Smart Context — Claude loads relevant memories based on CWD and git branch.

Decisions that stick — Authoritative memories that supersede conflicting old ones.

Session Handovers — Pass context to the next session seamlessly.

Contradiction Detection — Flags conflicting stored facts automatically.

Nightly Self-Learning — Validates memories against web sources overnight.

MEMORY.md vs BrainDB

Feature	MEMORY.md	BrainDB
Search	None	FTS5 + embeddings
Capacity	200 lines	Unlimited
Validation	None	Nightly fact-checking
Multi-agent	File conflicts	Claims + handovers
Contradictions	Manual	Auto-detected

The 51 Tools grouped

Memory (15): remember, recall, forget, decide, context, validate, learn...
Search (5): semantic-search, hybrid-search, embeddings...
Agents (8): heartbeat, claim, release, handover, inbox...
Graph (10): entities, relations, contradictions...
System (8): health, stats, backup, capabilities...
Inception (5): dream, decay, inception findings...

Every tool Claude Code needs for real persistent memory. No more MEMORY.md limitations.

GitHub: beckfexx/BrainDB — AGPL-3.0, TypeScript.

DEV Community