Your AI agent forgets everything between sessions.
CodeFire gives AI coding tools persistent memory, task tracking, and project intelligence. A native macOS companion for the AI-native workflow.
Works with Claude Code, Gemini CLI, Codex CLI, and OpenCode
The problem
You spend the first five minutes of every AI session re-explaining context. What you were working on. What files are involved. What decisions you already made. Your agent has no idea.
$ claude
You: Pick up where we left off on the auth refactor
Claude: I don't have context about a previous session or auth refactor. Could you tell me:
- Which files are involved?
- What changes have been made?
- What's the desired end state?
// 5 minutes of re-explaining later...
$ claude
You: Pick up where we left off on the auth refactor
Claude: Resuming auth refactor from your last session:
✓ Migrated to JWT tokens (src/auth/jwt.ts)
✓ Updated 3 of 7 route handlers
→ Remaining: payments, webhooks, admin
! Note: Rate limiter needs JWT update too
Continuing with the payments handler.
Features
A native macOS app that sits alongside your terminal, giving your AI agent the context and tools it needs to be genuinely useful.
Tasks, notes, and session context that survive across sessions. Your AI picks up exactly where the last conversation ended — no re-explaining required.
Built-in terminal, browser automation, and semantic code search. Your agent can run commands, navigate the web, and search your codebase by meaning — all through MCP.
One memory layer for every AI coding tool. Claude Code, Gemini CLI, Codex CLI, OpenCode — same persistent context, same project intelligence, any agent.
Kanban boards, priorities, labels, and task notes. Create tasks in one session, pick them up in the next.
Navigate, click, fill forms, and take screenshots. A headless browser controlled entirely by your AI agent.
Search by meaning, not just text. Indexes functions, classes, docs, and git history into a vector store.
Pull requests, commits, issues, and reviews. See your GitHub activity alongside your tasks and sessions.
Ask questions about your codebase with full project context. RAG-powered answers without leaving the app.
Pin architecture decisions, capture gotchas, get AI-generated daily briefings that summarize project activity.
See it in action
A native macOS app that puts your entire AI workflow side-by-side with your terminal.
Integrated terminal alongside the task board
Session history with cost tracking and tool usage stats
Task launcher, dev tools, and per-project cost tracking
Git integration with commits, staged changes, and diffs
Per-project notes with pinning and search
Built-in AI image generation
Getting started
Grab CodeFire.app from GitHub Releases and drag it to your Applications folder.
Click one button to install the MCP server for your CLI tool. No config files to edit.
Add your OpenRouter API key in Settings to unlock AI chat, code search, and image generation.
Your AI agent now has persistent memory, browser tools, code search, and more — across every session.
One-click setup
Click to install the CodeFire MCP server. Your CLI tool will connect to CodeFire automatically.
Don't have CodeFire installed? Download it first.
Prefer to configure things yourself? Add the MCP server manually.
$ claude mcp add codefire ~/Library/Application\ Support/CodeFire/bin/CodeFireMCP
// ~/.gemini/settings.json { "mcpServers": { "codefire": { "command": "~/Library/Application Support/CodeFire/bin/CodeFireMCP", "args": [] } } }
# ~/.codex/config.toml [mcp_servers.codefire] command = "~/Library/Application Support/CodeFire/bin/CodeFireMCP" args = []
// ~/.opencode/config.json { "mcpServers": { "codefire": { "command": "~/Library/Application Support/CodeFire/bin/CodeFireMCP", "args": [] } } }
Configuration
CodeFire uses external APIs for its AI-powered features. Add your keys in Settings (Cmd + ,).
Powers the built-in AI chat, semantic code search, and image generation. All models are served through OpenRouter with a single key.
sk-or-v1-...
Enables Gmail integration for email-to-task sync. Requires a Client ID and Client Secret from the Google Cloud Console.
Client ID + Client Secret
Built by developers who got tired of re-explaining context to their AI agents. MIT licensed. Contributions welcome.