Overview
Free Claude Code — FCC — is a local proxy and admin UI that sits between a coding agent CLI and whichever model provider you want it to talk to. It advertises 50 providers the project describes as terms-of-service friendly, covering free tiers, paid APIs, subscriptions and local servers, and states that it removes an integration if the provider stops permitting it. The README is explicit that the project is independent and not affiliated with or endorsed by Anthropic.
One installation serves ten agent CLIs — Claude Code, Codex, Pi, OpenCode, Cline, Hermes, DeepSeek Harness, Grok Build, Muse Code and Aider — each launched through its own wrapper command (`fcc-claude`, `fcc-codex`, `fcc-pi`, and so on) so the agent starts already pointed at the proxy. Agent behaviour is meant to survive the detour: streaming, tool calls, native interleaved thinking, image input, and independent routing of the Fable, Opus, Sonnet and Haiku slots to different compatible models.
Two other features are aimed at long sessions rather than at cost. When a provider exhausts its retries, FCC moves the turn to the next configured model instead of making you restart it, and it can cut terminal-output tokens using the optional RTK filter plus five built-in optimisations that answer quota probes, command-prefix detection, titles, suggestions and filepath lookups without calling a provider at all. Free-tier availability and limits belong to each provider and can change.
What it does
- One local proxy in front of 50 providers — free tiers, paid APIs, subscriptions and local servers — chosen from a searchable admin UI
- Ten supported agent CLIs, each with its own launcher: Claude Code, Codex, Pi, OpenCode, Cline, Hermes, DeepSeek Harness, Grok Build, Muse Code and Aider
- Automatic failover to the next configured model when a provider's retries are exhausted, without restarting the turn
- Optional RTK terminal-output filtering plus five built-in optimisations that answer routine agent requests locally
- Agent capabilities preserved end to end: streaming, tool use, native interleaved thinking, image input, and per-slot routing for Fable, Opus, Sonnet and Haiku
- Browser sessions for Codex with real-time and background runs, plus access from VS Code, JetBrains, the Codex App, Discord and Telegram
- Voice input through local Whisper or NVIDIA NIM transcription
- Optional bearer-token authentication on the local proxy
Getting started
FCC installs with a single script that also offers to install the agent CLIs you want and the optional RTK filter. Re-running the same script updates it.
Install or update
Run the installer for your platform. On Windows use the PowerShell form. Both scripts can be read in the repository before you run them.
curl -fsSL "https://raw.githubusercontent.com/Alishahryar1/free-claude-code/main/scripts/install.sh" | shStart the server
On macOS and Windows, open Free Claude Code from the Applications folder or Start menu and use the tray icon to reach Admin. On Linux, start the server from a terminal and keep it open.
fcc-serverConfigure a provider
FCC opens the Admin UI once the server is running. Create an API key at build.nvidia.com/settings/api-keys, paste it into NVIDIA_NIM_API_KEY, leave MODEL on the default or pick another from the dropdown, and click Apply. Enable Proxy Authentication if you want a bearer token in front of the local proxy.
nvidia_nim/nvidia/nemotron-3-super-120b-a12bLaunch an agent through FCC
Each supported CLI has its own wrapper command, which starts the agent already routed through the proxy.
fcc-claude
fcc-codex
fcc-pi
fcc-opencode
fcc-cline
fcc-hermesCommands and code are distilled from the project's own documentation — always check the official repo for the latest.
When to use it
- Reach for it when you want to keep using a coding agent CLI without an API key from that agent's vendor
- Reach for it to keep a long session alive across provider outages, since a failed turn moves to the next configured model instead of restarting
- Reach for it when a local llama.cpp or LM Studio server should back the same CLI you use against hosted providers
- Reach for it to cut terminal-output tokens on agent runs that read a lot of command output
How Free Claude Code compares
Free Claude Code alongside other open-source gateways & routing tools AI/TLDR tracks, ranked by GitHub stars.
| Tool | Stars | What it does |
|---|---|---|
| OmniRoute | ★ 64.8k | A local-first AI gateway that fronts hundreds of model providers with one OpenAI-compatible endpoint, quota-aware auto-fallback, many routing strategies and prompt compression. |
| LiteLLM | ★ 58.5k | A Python SDK and proxy server that gives one OpenAI-compatible API to 100+ LLM providers, with cost tracking, budgets, fallbacks, rate limiting, and an admin UI. |
| Free Claude Code | ★ 54.5k | Point Claude Code and nine other agent CLIs at free, paid or local model providers |
| Kong Gateway | ★ 44.1k | A Lua/NGINX API gateway whose AI Gateway plugins put one universal endpoint in front of OpenAI, Anthropic, Gemini, Bedrock and other providers, with semantic routing, caching and MCP traffic governance. |
| Apache APISIX | ★ 17.1k | A cloud-native API gateway whose AI plugins add multi-provider LLM proxying, load balancing, retries and fallbacks, token-based rate limiting, and content moderation. |
| Portkey AI Gateway | ★ 13k | An LLM gateway that routes calls to 100+ providers through one API and adds logging, tracing, caching, and fallbacks for production AI traffic. |
| Higress | ★ 9.4k | An AI-native API gateway built on Istio and Envoy that proxies and governs traffic to many LLM providers, with token rate limiting, caching, and MCP server hosting. |
| Bifrost | ★ 8k | A high-throughput LLM gateway written in Go that gives a single OpenAI-compatible API to many providers, with failover, load balancing, semantic caching, and very low overhead at high request rates. |