← Portfolio/free-code

free-code

Your AI coding agent. Runs natively on macOS, in VS Code, and in the terminal.

A full-featured AI coding agent with deep code understanding, local knowledge bases, browser automation, multi-provider AI, and layered security — local-first, cloud-ready.

Linux & macOS VS Code / Cursor plugin Terminal CLI macOS desktop app Anthropic · Vertex · Ollama · LM Studio
3 Platforms
6+ AI Providers
40+ Editor Commands
100% Local & Private

What makes free-code different

High-level principles — how the agent behaves, where it runs, and how teams trust it.

Agentic

Real tools on your project: read and edit files, run shell commands, browse the web, and call MCP servers — it acts, not only suggests.

Local-first

Sessions, code graph, and knowledge bases live on your machine. Your source stays under your control by default.

Extensible

Skills, extensions, and MCP servers extend behavior without forking the agent. Install once, share with the team via version control.

Efficient

Code graph and structured context cut wasted tokens — the agent reads less, sends smaller prompts, and iterates faster instead of scanning whole files.

Connected

Reaches beyond your repo: calls MCP servers, receives inbound webhooks from external systems (Flowise, CI, GitHub), and explores the web autonomously — browsing pages, searching, and fetching what it needs to finish a task.

External integrations →

Secure

Layered guardrails: export conversations for audit, confirm risky bash commands in real time, see which skills, MCPs, and agents are active, and deploy locally or in the cloud with policies your team controls.

Security layers →

Built-in capabilities

Specific tools and integrations included with free-code — what you can use in a session.

Multi-Provider AI

Anthropic, Google Vertex, Ollama, LM Studio, or any OpenAI-compatible endpoint. Switch models mid-session with Ctrl+L or cycle a curated shortlist with Ctrl+P.

Full Tool Access

Read, edit, and write files. Run shell commands. Search with grep and find. The agent works on your actual project — not a sandbox copy.

RAG Knowledge Base

Local vector search over your own documents. Index files, folders, or entire GitHub repos. Multiple named knowledge bases. Auto-refresh on a schedule.

Browser Automation

Built-in browser control with CDP. Navigate pages, take screenshots, extract content, and interact with web UIs. Gemini integration included.

MCP Servers

Connect any Model Context Protocol server — databases, APIs, CI systems, design tools. One import command and the tools appear in the next session automatically.

Inbound Webhooks

The reverse of MCP: external systems — Flowise flows, CI, GitHub — POST events into a running session. Queue them for the agent to drain, or trigger a turn immediately. Local, authenticated, one auto-assigned port per session.

How webhooks work →

Subagents

Spin up parallel subagents from within a session. Each runs independently with its own context. Tackle large tasks without a single context window.

Skills & Extensions

Commit reusable agent behaviors as markdown skills alongside your code. Extensions add new tools and commands — install in one line, shared with the team.

Session Management

Resume, fork, compact, and branch sessions. Navigate history as a tree. Export to HTML, Markdown, or JSONL for sharing or archiving.

Code Graph

Local symbol and call-edge index. Find any function by name, trace callers, and read symbol source — without blindly scanning files.

See it in action

Interactive mocks of both interfaces. Try typing / in the input to open the command menu.

Quick start

From zero to a running session in under two minutes.

Runs on Linux and macOS
The installer sets up the CLI (on macOS, also the desktop app). The VS Code / Cursor plugin is a quick separate step you do by hand afterwards. Both installers live in the installation/ folder.

1. Fast installation

Run the script for your platform from the installation/ folder:

# Linux — installs the CLI
bash ./installation/install-free-code-linux.sh

# macOS — installs the CLI + desktop app
bash ./installation/install-free-code-mac.command

On macOS you can also double-click install-free-code-mac.command in Finder.

Want the editor plugin too? See Install the VS Code / Cursor plugin.

2. Start a session

free-code

On macOS you can also open the FreeCodeMac desktop app from /Applications.

Why free-code?

What sets it apart from generic AI assistants.

An agent, not a chatbot

free-code has real tools: it reads and edits files, runs shell commands, controls a browser, queries knowledge bases, and calls MCP servers. It doesn't just suggest — it acts.

Local-first, cloud-ready

No cloud indexing of your repo by default. Sessions, knowledge bases, and the code graph are local files you control. Use cloud LLM providers when you want, or run in managed environments — with security layers such as Damage-Control and MCP authorization where your policy requires them.

Your AI provider, your choice

Swap between Claude, Gemini, GPT, Llama, Mistral, or any local model without changing your workflow. One key binding cycles your personal shortlist.

Not just a chatbot

The agent has full access to your filesystem, terminal, browser, and any MCP server you connect. It reads, edits, runs tests, checks CI, and follows up — without you babysitting it.

Built for teams

Skills are plain markdown files you can commit alongside your code. Share domain knowledge, onboarding flows, and repeatable workflows with your team through version control.

Sessions that survive context limits

Compact long conversations without losing thread. Fork sessions to explore alternatives. Resume after days. Navigate history as a tree and jump to any checkpoint.