Star 历史趋势
数据来源: GitHub API · 生成自 Stargazers.cn
README.md

FreeRide

The OpenAI-compatible gateway for every free-tier provider.

freeride-now-supports-claudecode

One local endpoint that fans out across OpenRouter, Groq, NVIDIA NIM, HuggingFace, Cerebras, Cloudflare Workers AI, and your own Ollama. Hit a rate limit, fail over to the next provider. Your agent never knows.

Also wraps Claude Code, OpenAI Codex, and Google Gemini CLI — run all three without paying any of their vendors.

102M+ tokens served in 35 days. $0 spent. Routed through community free-tier keys via this gateway. Daily traffic: free-ride.xyz/models

curl -sSL https://api.free-ride.xyz/install.sh | sh
freeride run claude    # or: freeride run codex / freeride run gemini

That's it. No accounts, no subscriptions, no FreeRide cloud. Local-first, BYO keys, your machine talks to providers directly.


Install

macOS / Linux:

curl -sSL https://api.free-ride.xyz/install.sh | sh

Windows (PowerShell):

powershell -ExecutionPolicy ByPass -c "irm https://api.free-ride.xyz/install.ps1 | iex"

The installer picks up uv, then pipx, then plain pip — whichever is on your system. Or install from source.

freeride init           # interactive — collects keys, writes ~/.freeride/.env
freeride serve          # gateway listens on localhost:11343

Get keys (any one is enough; more = better failover):

ProviderFree tierGet a key
OpenRouterrotating free modelsopenrouter.ai/keys
Groqdaily token capconsole.groq.com/keys
NVIDIA NIMcredits per accountbuild.nvidia.com
HuggingFace$0.10/mo Free, $2/mo PROhuggingface.co/settings/tokens
CerebrasRPM / TPM capscloud.cerebras.ai
Cloudflare Workers AI10K neurons/daydash.cloudflare.com
Ollama (local)no quotainstall from ollama.com

Run a coding agent — free

Three of the major coding CLIs ship a freeride run wrapper that works with no per-vendor key and no login. The gateway translates between each CLI's native wire protocol and our routing layer; you get the polished agent UX of each CLI, paid for entirely by free-tier providers.

Claude Code

freeride run claude

Inside the session, switch routing per request via /model:

You typeWhat happens
/model claude-opus-4-7Your Pro/Max subscription answers (passthrough to api.anthropic.com) — only if claude login has run
/model freeride/freeFree providers answer; smart-router picks the model
/model freeride/fastFree; prefers Groq (low TTFT)
/model freeride/qualityFree; prefers OpenRouter (widest catalog)
/model freeride/codingFree; pinned to a code-tuned model that reliably emits tool_use blocks

Full guide: docs/agents/claude-code.md.

OpenAI Codex

freeride run codex

Whatever model the CLI picks (gpt-5-codex, gpt-5, etc.) is routed to a free upstream provider. The gateway translates the Responses-API wire format (with full SSE event protocol — response.output_item.addedoutput_text.deltaoutput_item.doneresponse.completed) so the CLI parses everything natively.

Note: codex uses bubblewrap for shell-tool sandboxing; on systems without it, file/shell tool calls fail (the model still works). Full guide: docs/agents/codex.md.

Google Gemini CLI

freeride run gemini

Any gemini-* model name routes to a free upstream provider. Translator handles Google's {contents, tools, generationConfig} shape both directions. Full guide: docs/agents/gemini.md.

Any other agent / SDK

# Aider / Continue.dev / hermes / your-own-tool — anything that speaks OpenAI:
freeride bind aider
freeride bind continue
# or just point it at the gateway directly:
OPENAI_API_BASE=http://localhost:11343/v1
OPENAI_API_KEY=any-string-here

How failover works

Per-request the chain is (provider, key), sorted by recent health:

  1. Try the head pair.
  2. RATE_LIMIT or AUTH error → mark the key as cooling, try the next key on the same provider.
  3. MODEL_NOT_FOUND or QUOTA_EXHAUSTED → skip to the next provider.
  4. 5xx / TIMEOUT → next pair.
  5. First successful response — stamp X-FreeRide-Provider + X-FreeRide-Request-Id headers and ship.

If every pair fails, you get a structured 503 with a per-provider breakdown so debugging is one log line, not five round-trips. Mid-stream errors after the first chunk shipped are logged but don't break the client (we can't un-ship bytes).

Smart routing for model: "auto": the resolver scores every free model in the catalog by health × popularity (from the public models leaderboard) and picks the best one. Run freeride audit-models once after install to cache health probes locally so the first real request isn't a cold start.

Deeper: docs/architecture/failover.md.


Providers

ProviderSurfaceNotes
OpenRouterchat, streaming, tools, vision, structured outputs, embeddingsfull surface — the most-used provider in our routing
NVIDIA NIMchat + embeddingscurated free-model allowlist; NVIDIA_NIM_FREE_MODELS_OVERRIDE to expand
GroqchatLlama 3.x, Gemma 2, Mixtral, DeepSeek-R1-distill; daily token cap
Cloudflare Workers AIchatcheap-per-neuron models; needs CLOUDFLARE_ACCOUNT_ID
HuggingFace Inferencechat + embeddingsfull HF router catalog; budget governs access
Cerebraschatfastest Llama / Qwen inference; no embeddings
Ollama (local)chatlocal-only; can mix with remote in the same failover chain

Adding a new provider: implement freeride.core.provider.Provider in freeride/providers/<name>.py, register it in the conformance suite. See CONTRIBUTING.md.


Multi-key rotation

Provide more than one key per provider with a numbered suffix:

OPENROUTER_API_KEY=sk-or-v1-aaa     # primary
OPENROUTER_API_KEY_2=sk-or-v1-bbb
OPENROUTER_API_KEY_3=sk-or-v1-ccc

The router tries them in health order. A 429 on one key cools it for the next 60s and rotates to the sibling key — no provider switch needed. On startup freeride keys shows which keys are available vs cooling.


See what the gateway is doing

freeride doctor                # static checks: keys, ports, /etc/hosts, common gotchas
freeride doctor --claude-code  # the same + Claude-Code-specific probes
freeride audit-models          # probe every free model on every key; cache the results
freeride bench                 # measure p50/p95/tok-s per provider

Tail live events:

tail -f ~/.freeride/events.jsonl

Each line is a JSON event: routing decisions, provider attempts, response statuses, mid-stream errors. Same schema the marketing site reads to render the live token counter and provider leaderboard.


Telemetry

A small beacon ships hourly with counts only: tokens served, request count, active providers, uptime hours, OS, version, and a per-install UUID. Never sent: prompts, completions, model IDs, API keys, hostname, IP.

freeride telemetry        # audit what the next beacon would post
freeride telemetry off    # opt out

The aggregate is what powers free-ride.xyz/models. Default on; explicit disclosure banner prints on first run.


Commands

freeride init           interactive setup wizard — prompts for keys, writes ~/.freeride/.env
freeride serve          start the gateway on :11343
freeride run <cli>      wrap a CLI (claude / codex / gemini / anything) — points it at the gateway
freeride bind <agent>   write the agent's config so it uses the gateway permanently
freeride doctor         pre-flight checks: keys, ports, hosts file, common gotchas
freeride keys           which provider keys are available vs cooling
freeride audit-models   probe every free model; cache health locally
freeride bench          measure p50/p95/tok-s per provider
freeride list           list available free models
freeride telemetry      manage the hourly aggregate beacon

Docs


License

MIT.

关于 About

One free AI endpoint, every free tier behind it. Local OpenAI-compatible gateway routing across OpenRouter, Groq, NVIDIA NIM, Cloudflare Workers AI, and HuggingFace with automatic failover.
ai-gatewayclaudeclaude-codeclicloudflare-workers-aifailoverfree-tiergroqhermeshuggingfacellmnvidia-nimopenaiopenai-compatibleopenclawopenrouterpython

语言 Languages

Python95.2%
JavaScript3.0%
Shell1.3%
PowerShell0.5%

提交活跃度 Commit Activity

代码提交热力图
过去 52 周的开发活跃度
180
Total Commits
峰值: 144次/周
Less
More

核心贡献者 Contributors