What is CodeFire?
CodeFire is a native macOS companion app for AI coding agents like Claude Code, Gemini CLI, Codex CLI, and OpenCode. It solves a fundamental problem: your AI coding agent forgets everything between sessions. CodeFire creates a persistent memory layer that auto-discovers your projects, tracks tasks and sessions, monitors live coding activity, and exposes project data back to your AI via MCP — creating a feedback loop where your agent knows what you’re working on and can act on it.Key Features
Project Management
Auto-discovers Claude Code projects from
~/.claude/projects/. Each project opens in its own window with an integrated terminal and tabbed GUI panel.Task Tracking
Drag-and-drop Kanban board (Todo / In Progress / Done) with priority levels, labels, notes, and full task history. Tasks can be created manually, from emails, or programmatically via MCP.
Live Session Monitoring
Real-time mission control for active coding sessions. Track token usage, costs, tools invoked, files touched, and a live activity feed.
Session History
Browse past conversations, review tool usage patterns, and track costs over time with full session parsing and indexing.
Built-in Terminal
Tabbed terminal emulator (SwiftTerm) embedded in each project window. Launch sessions, run commands, manage multiple tabs.
GitHub & Git Integration
View commits, staged changes, diffs, open PRs, and CI status without leaving the app.
Notes & Context
Per-project and global notes with rich editing, pinning, and search to persist context between sessions.
AI-Powered Features
Built-in AI chat, semantic code search, and image generation powered by OpenRouter.
MCP Server Integration
CodeFire includes a companion MCP server (CodeFireMCP) that exposes your project data to any AI coding tool. When configured, your agent can:
- List and manage tasks (create, update status, add notes)
- Read project notes and search across them
- Access the codebase profile and file tree
- Query session history
- Navigate and interact with web pages
- Generate images
This creates a powerful feedback loop: you manage work in CodeFire, and your AI has full awareness of that work during coding sessions.
Architecture
CodeFire is built with Swift Package Manager and consists of two executable targets:| Target | Description |
|---|---|
CodeFire | Main GUI app — SwiftUI + AppKit, no Xcode project needed |
CodeFireMCP | Standalone MCP server binary, communicates via stdio |
Key Dependencies
- GRDB.swift — SQLite database shared between app and MCP server
- SwiftTerm — Terminal emulator
- WhisperKit — Speech recognition capabilities
No Electron. No web views (except the built-in browser). No node_modules. The entire app is approximately 16MB.
Data Storage
All data lives in a single SQLite database at~/Library/Application Support/CodeFire/codefire.db, shared by both the GUI app and the MCP server.
Requirements
- macOS 14.0 (Sonoma) or later
- One or more supported CLI tools:
- GitHub CLI (
gh) for GitHub integration (optional)
Next Steps
Quick Start
Get CodeFire up and running in minutes