Quick Start

Get Cortexa running and capture your first thought in under 5 minutes.

1. Install

Docker is not required. Cortexa does not bundle, start, or depend on a container engine for the desktop app, vault, Sources, Foundry, MCP, Hive, or agent-memory features. Container instructions elsewhere in the repository apply only to an optional maintainer evaluation harness.

Windows

Download the installer from GitHub Releases: _x64-setup.exe for Intel/AMD or _arm64-setup.exe for ARM. Runs per-user, no admin needed.

Build from Source

git clone https://github.com/cortexalabs/cortexa.git
cd cortexa
npm install
npm run tauri dev    # dev mode with hot reload
npm run tauri build  # release build

Requirements: Rust 1.77+, Node.js 18+, and platform-specific deps (see Installation).

2. Set Up an LLM Provider

Cortexa needs at least one chat provider (for summaries/thesis) and one embedding provider (for semantic search).

Quickest path — Ollama (free, local):

ollama pull llama3.2          # chat (~2 GB)
ollama pull nomic-embed-text  # embeddings (~270 MB)

Or use cloud providers (Anthropic, OpenAI, Google Gemini, Voyage) via API key in Settings → LLM.

3. Capture Your First Thought

Press Ctrl+Shift+P (default hotkey). A capture window appears. Type something:

/idea: Use Cortexa as persistent memory for my AI coding agents

Press Ctrl+Enter to save. The thought is:

1. Encrypted with AES-256-GCM
2. Queued for embedding (semantic vector)
3. Auto-tagged with idea (from the /idea: template)
4. Queued for Deep Think analysis (the template's default action)

4. Connect Your AI Agent

Claude Code (recommended)

claude mcp add --transport http cortexa http://127.0.0.1:7373/mcp \
  --header "Authorization: Bearer <TOKEN>"

Get the token from Settings → External in the Cortexa app.

Other agents

See Integrations for Cursor, Codex, Gemini, Claude Desktop setup.

5. Use It

In your next Claude Code session:

# Agent calls cortexa_context automatically or you ask it to:
"Check Cortexa for any prior decisions about caching"

Or generate a context file that all agents read automatically:

cortexa-cli sync-context MyProject --dir /path/to/project

This creates .cortexa-context.md with your relevant decisions, patterns, and TODOs — read by Claude Code, Cursor, and any agent that scans the project directory.

What's Next?

• Thoughts & Captures — capture types, metadata, and session grouping
• Search — hybrid semantic + keyword search with context boosting
• Deep Think — the expand → critique → synthesize pipeline
• Context Injection — how agents consume your vault
• MCP Tools Reference — every MCP tool explained