- Home
- Skills
- Agents & Orchestration
- Global Brain: Cross-Project Claude Code Memory
Works with the AI Agents you already use
Global Brain: Cross-Project Claude Code Memory
Give Claude Code one memory across all your projects. A local SQLite brain captures commits, topics and files from your transcripts and loads a small digest into every session.
Free
Global Brain: Cross-Project Claude Code Memory
Example session with this skill installed
/global-brain what did we decide about pagination?
- Read your context and instructions
- Compiled the global brain
Real CLI output from a test install, with sample entries:
1 result — "pagination":
#1 [api · decision · 2026-11-04] 🌐 Use cursor pagination
Offset pagination timed out on large tables; all list endpoints use cursors.
(imp 70 · manual)
The agent then answers in prose and cites the entry. The digest loaded into new sessions:
## 📌 Cross-project / pinned
- [api · decision · 2026-11-04] Use cursor pagination — Offset pagination timed out on large tables; all list endpoints use cursors.
## web
- [web · preference · 2026-11-04] Dates in UTC — Store dates in UTC and format them only in the UI.
Connects securely to your tools. The creator never sees your data.
About this skill
Claude Code starts each project without what you learned in the others. Global Brain keeps that knowledge in one local database and loads a size-limited digest into every session.
- Automatic capture: reads Claude Code transcripts and extracts commits, topics, files and goals.
- Saved on request: store decisions, preferences and gotchas with
remember, and search them withquery. - Bounded: token and per-project limits keep the digest small.
- Local: memory stays in
~/.claude/global-brain/brain.db. - Secret-aware: values that look like API keys, tokens or passwords are masked as
[REDACTED]before they are stored or loaded into the digest. - Clean removal:
global-brain uninstallremoves the hooks, restores your previous status line and removes theCLAUDE.mdimport. Your memory is kept unless you pass--purge. - Open source: MIT licensed, with zero runtime dependencies.
What setup changes: this skill installs the @kxrk0/global-brain npm package after asking you. It adds two hooks, wraps your status line (the previous one is saved and restored on uninstall), and adds one import line to CLAUDE.md. It also checks npm for updates in the background, which you can turn off.
Works only with Claude Code and requires Node 22.13+. Automatic capture is heuristic (no LLM) and can miss details, so use remember for anything important. Secret masking is pattern-based: it catches common key and token formats and password=-style values, but do not paste secrets into prompts on purpose. Background version checks run at most every 15 minutes unless disabled with GLOBAL_BRAIN_NO_UPDATE_CHECK=1.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
Trust & safety
Security scan
Security scan passed 12 days ago
- Free with an Agensi account
Requires
Permissions
File Scopes
Shell runs the global-brain CLI and the npm install, after user approval. File access covers ~/.claude/global-brain, ~/.claude/skills/global-brain, settings.json, CLAUDE.md, global-brain.md and the transcripts in ~/.claude/projects (read only). Network use is the npm install and a version check against registry.npmjs.org. The database stays on your machine; the digest is loaded into Claude Code sessions like any CLAUDE.md content, with values that look like secrets masked first. `global-brain uninstall` reverts the settings.json and CLAUDE.md changes.