Stop juggling separate signups and keys for every AI provider. Sign in once, create a key, call one endpoint — and pick any model per request, from fast free models to the world's most advanced models.
GitHub accounts must be at least 60 days old to reduce abuse from throwaway accounts.
Frontier models, ranked by capability. Every one of them is available with Buddy Premium — no five separate subscriptions.
Longer context = the whole book in one prompt. Ours span from 128K to 1.05M tokens.
Same endpoint every time — just change the model field to switch. Live catalog below.
One account, one key. No credit card, no email spam — just a 60-day-old GitHub account.
Name it anything, set your own daily quota (or leave it unlimited), choose an expiration.
OpenAI-compatible format. Swap the model field per request — free, premium or paid.
Skip separate accounts and keys for every model provider — one key covers all of them.
Same endpoint every time. Just change the "model" field to switch.
Claude Opus 4.8, GPT-5.6 Sol, Gemini 3.1 Pro, GPT-5.5 Pro, Claude Fable 5, Grok 4.5 & more — try them free before subscribing.
Every signed-in account gets a free trial pool of premium messages — no card required to taste the frontier.
GitHub-only sign-in, account-age checks, and bot verification keep it usable for everyone.
OpenAI-compatible request format — drop-in for tools that already speak that format.
They switched to one key and never looked back.
“One key for every model changed my workflow completely. The GPT-5.6 Sol model is genuinely impressive.”
“The free premium trial sold me instantly. Clean API, no Bearer hassle — and the frontier models punch way above the price.”
“We route our whole prototype through Buddy. 60€/year for this lineup is a no-brainer, and the chat UI is gorgeous.”
No "Bearer" prefix needed — the bare key in the Authorization header works.
curl -X POST https://buddy-router.pages.dev/api/v1/chat/completions \
-H "Authorization: YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai/gpt-oss-20b",
"messages": [{"role": "user", "content": "hi"}]
}'
Sign in to see your keys, the full model list and its specs, and ready-to-use snippets in cURL, Python, JavaScript and Node.js.
Yes — standard access doesn't cost anything. Free models are unlimited-ish (you choose your key's daily quota). Premium models add a 6€/month or 60€/year option, with a free trial before you pay anything.
They're Buddy's frontier-tier models — Claude Opus 4.8, Claude Fable 5, GPT-5.6 Sol, GPT-5.5 Pro, Gemini 3.1 Pro, Mistral Large 3, Grok 4.5, Llama 4 and Kimi K3. Each has its own personality and strengths. Subscribe for unlimited use, or spend your free trial pool first.
It's a simple anti-abuse gate. It stops the service from being drained by bots spinning up throwaway accounts, without requiring any personal info beyond your existing GitHub profile.
Each key gets a daily request quota you choose when creating it (leave blank for unlimited), resetting at midnight UTC. You can hold up to 3 active keys. The premium trial pool is a one-time budget of 10 premium messages per account.
Premium is billed securely through PayPal (6€/month or 60€/year). You can manage or cancel your subscription anytime from your PayPal account.
Yes — that's the plan. We intend to open-source the Buddy Router core (the API gateway) so the community can self-host, audit and extend it. We'll announce it on the project repo as soon as it's ready — no date locked in yet, but it's a real roadmap item.
Send a POST to /api/v1/chat/completions with the key in the Authorization header — the bare key works, no Bearer prefix needed (a legacy Bearer is still accepted). The body is OpenAI-compatible: a model field plus a messages array. Set "stream": true for streaming (SSE) responses.
Prompts are used only to serve your request and are not used to train anything. API keys are stored hashed, so even we can't read back a raw key after creation. Some upstream models (DeepSeek) flag that they may use prompts for training — those models carry a visible "trains" badge so you can avoid them.
Yes — every model accepts text and code files (.txt, .md, .json, .py, .js, .ts, …). Images are only accepted by vision-capable models (they're marked 👁 in the model list); non-vision models silently drop images.
Yes — there's a 🧠 toggle in the chat composer. With thinking off, reasoning models answer faster without chain-of-thought. Some models (e.g. Buddy 5.0) are fast enough that you won't notice a difference.
All Buddy-tier models (Buddy 1.0 through 5.0) are free with your account, each with its own daily quota shown in the model picker. The raw gateway models (openai/gpt-oss-*, qwen/*, minimax/*, z-ai/glm-5.2, …) are also free via the API with your key's quota.