For developers using AI daily

Your AI forgets everything.
Cortex remembers.

Every new AI session starts from zero. You re-explain your stack, your decisions, your project structure — every single time.

Cortex gives Claude, Cursor, and Copilot a persistent memory layer connected to your GitHub repo. The AI reads everything before it responds, and writes files directly as real commits.

No credit card · Works with Claude & Cursor
🔓 Open source 🔒 Your code stays in your own GitHub repo ⚡ Set up in under 2 minutes
Claude Desktop — cortex connected
You: add a dark mode toggle to the settings page
Claude: ✓ read_context() — called automatically, no prompt needed
→ Read settings.jsx, theme.js, cortex.context
→ Saw previous decision: using CSS variables for theming
→ Stack: React + Tailwind, tone: clean & minimal
✓ save_progress("settings.jsx") — committed to GitHub, history updated
Done. Used your existing --color-bg variables and matched
the toggle style from your ProfilePage component.

Cortex connects to your AI tools using MCP (Model Context Protocol) — the open standard that lets AI tools securely read and act on outside data. Cortex uses it to give every tool you use the same shared memory.

Not the same as your tool's built-in memory. Cursor and Copilot only remember within their own app. Cortex's memory lives in your GitHub repo — so Claude, Cursor, and Copilot all share the exact same context, automatically.
Step 1
Create a Space
Describe your project once. Cortex creates a cortex.context file in your GitHub repo — the AI's permanent memory that evolves over time.
Step 2
Connect your AI tool
One config paste into Claude Desktop or Cursor. The AI gets 5 tools: read files, write files, update memory. Works with VS Code Copilot and Windsurf too.
Step 3
Never explain twice
AI reads your full context before every response. Every file it writes goes to GitHub as a real commit. Every change is tracked in cortex.map.
Free to use · Open source · Built for developers who use AI every day
Spaces
⬡

Welcome to Cortex

Create a Space and connect your GitHub repo. AI will read and write your files automatically — with full project memory.

📋 Overview
🗺️ Cortex Map
🔌 Connect AI
Space Info
Name—
GitHub Repo—
Brand—
Industry—
Tone—
cortex.context View on GitHub ↗
Loading…
cortex.map auto-updated by AI
🔗 Your Personal MCP URL
This URL connects Claude Desktop or Cursor directly to your Space. It's unique to you — no downloads, no local setup.
Generating your URL…
🖥️ Claude Desktop Config
Replace the entire contents of your Claude Desktop config file with this. Requires Node.js installed (nodejs.org).
📁 Config file:
Windows: %APPDATA%\Claude\claude_desktop_config.json
Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
Loading…
!
Fully quit Claude Desktop (system tray → right click → Quit) then reopen. Look for 🔨 hammer icon.
✓
That's it — just talk normally. Claude reads your project automatically and saves changes to GitHub as it works.
⚡ Cursor Config
Create or open ~/.cursor/mcp.json and paste this:
Loading…
🤖 Connect Other AI Tools
Claude Desktop and Cursor are set up above. Click any tool below for its exact setup steps.
🪟 VS Code (GitHub Copilot)
VS Code uses a different config key (servers, not mcpServers) and requires Agent mode.
1
Create .vscode/mcp.json in your project, or run MCP: Open User Configuration from the Command Palette for a global setup.
2
Paste this:
Loading…
3
Open Copilot Chat → switch mode to Agent → click the 🛠️ tools icon to confirm Cortex tools are listed.
🌊 Windsurf
1
Open ~/.codeium/windsurf/mcp_config.json (Mac/Linux) or %USERPROFILE%\.codeium\windsurf\mcp_config.json (Windows). Use the gear icon in the Cascade panel → MCP Servers if you'd rather not find it manually.
2
Paste this:
Loading…
3
Click Refresh Servers in the Cascade MCP panel.
⌨️ Claude Code (CLI)
1
Run this in your terminal (inside your project folder):
Loading…
2
Start a session with claude as usual — Cortex tools will be available.
🪿 Goose
1
Run goose configure → Add Extension → Command-line Extension.
2
When asked for the command, paste:
Loading…
💬 ChatGPT
ChatGPT connects to remote MCP servers through its own UI rather than a config file, and currently expects either no auth or full OAuth — not a custom header — so the setup below is best-effort and may not expose write access the same way Claude/Cursor do. Requires a paid plan (Plus/Pro/Business) with Developer Mode enabled.
1
Settings → Connectors → enable Developer Mode if not already on.
2
Add custom connector → paste your MCP URL (shown at the top of this page) as the server URL. Leave auth as "none" — token-based header auth isn't supported in ChatGPT's connector UI yet.
3
In a new chat, click + → More → Developer Mode → select Cortex as a source.
⚠️ Claude.ai (web) does not yet support adding custom remote MCP servers as of this writing — only Claude Desktop and Claude Code do.
🛠️ Tools Available to AI
read_context()
Called automatically first — reads all files, cortex.context, and change history
save_progress(path, content, is_new)
One call saves the file to GitHub and updates the change history — no separate map call needed
ask_codebase(question)
Ask a targeted question — routed through Groq, returns a short answer instead of full file dumps to save tokens
read_file(path)
Read one specific file from GitHub
update_context(content)
Updates cortex.context when project goals or stack change
list_files()
Lists all files in the GitHub repo
delete_file(path)
Deletes a file from GitHub
⚡ Claude is told to call read_context automatically when it connects — you no longer need to ask it to.
1
Details
2
GitHub
3
Repository
Create a Space
A Space is a persistent AI memory for one project or client.
🏢 Brand Context (optional)