brain_portability

Model-Neutral AI Brain

"Bring your own model, keep your brain." cachly is the memory layer — not the model. Your lessons, crystals, predictions, and team knowledge live in your Brain instance, accessible from any MCP-compatible AI client. Switch editors tomorrow; your Brain is already there.

Why model-neutrality matters

Vendor lock-in
Memory stored inside the model provider (Anthropic, OpenAI) is inaccessible the moment you switch tools. Your team's hard-won knowledge vanishes.
cachly approach
Your Brain lives in a neutral Redis/Valkey store you control. Any MCP-compatible client can connect — same ID, same lessons, zero migration.

7 supported clients — one Brain

🤖Claude CodemcpServers
~/.claude/settings.json
🖱️CursormcpServers
~/.cursor/mcp.json
🏄WindsurfmcpServers
~/.windsurf/mcp.json
🐙GitHub Copilotservers
.vscode/mcp.json
ClinemcpServers
cline_mcp_settings.json
🔷Zedcontext_servers
~/.config/zed/settings.json
▶️ContinuemcpServers
~/.continue/config.json

One command — all editors

autopilot detects every installed editor and writes the correct config for each one — with your Brain ID pre-filled.

npx @cachly-dev/mcp-server@latest autopilot

Already configured one client? Run autopilot again — it adds the missing ones without touching your existing setup.

What travels with you across clients

DataModel-neutral?
Lessons (learn_from_attempts)✅ Full access in all clients
Memory Crystals (memory_crystalize)✅ Full access in all clients
Causal Knowledge Graph (brain_predict)✅ Full access in all clients
Team lessons (team_learn / team_confirm)✅ Full access in all clients
Domain Brain Marketplace installs✅ Full access in all clients
Session handoffs (session_handoff)✅ Full access in all clients
Cross-author collaboration graph✅ Full access in all clients
Model weights / chat historyNot stored by cachly

Verify with brain_portability()

Run brain_portability() inside any connected client. It returns your Brain ID and ready-to-paste config blocks for all 7 clients — the living proof that your Brain is not tied to the model you happen to be using right now.

brain_portability()

→ Brain ID: a3f7c...
→ 7 compatible clients
→ Config blocks for: Claude Code, Cursor, Windsurf,
    GitHub Copilot, Cline, Zed, Continue
→ Model-neutral data: lessons ✅ crystals ✅ predictions ✅