# ZeroLimitAI

> OpenAI-compatible AI API and chat platform. Send model: "auto" and ZeroOptimize™ routes each request to the best free model available right now (for example Google's Gemma; the free catalog changes week to week and is re-ranked daily), with automatic failover. Inference cost is $0 on every plan; plans differ by daily call volume.

## API

- Base URL: https://www.zerolimitai.com/api/v1
- Auth: Authorization: Bearer zlai_<key> — create keys at https://www.zerolimitai.com/developer
- Endpoints: POST /chat/completions (alias POST /chat), GET /models, GET /usage
- Request body: { "model": "auto", "messages": [{ "role": "user", "content": "..." }], "stream": false, "agentId": "optional" }
- Optional sampling: temperature (0-2), top_p (0-1), stop (string or up to 4), seed, presence_penalty and frequency_penalty (-2 to 2) are forwarded to the provider; max_tokens / max_completion_tokens (1-32768) caps the answer (finish_reason "length"). Out-of-range values: 400 invalid_param.
- Function calling: send OpenAI tools (tool_choice, parallel_tool_calls optional); routed to the best free model that supports tools, with failover; tool_calls returned in OpenAI format, streaming too. Tool results go back as role "tool" messages.
- Response: OpenAI chat.completion shape; the "model" field reports the model that actually answered
- Errors: { "error": { "message", "type", "code" } } — codes: invalid_key, upgrade_required, account_suspended, invalid_json, missing_messages, invalid_param, agent_not_found, request_too_large (free accounts, over ~16,000 input tokens), rate_limited, daily_limit, upstream_error
- Limits: 60 requests/min per key; daily quota in X-RateLimit-Limit / X-RateLimit-Remaining / X-RateLimit-Reset headers
- Works unchanged with: openai (Python/Node), LangChain, Cline, Roo Code, Continue, Aider, Open WebUI — set base_url and api_key, use model "auto"
- Not available: embeddings, image inputs on the API
- Privacy: we don't train on requests; requests only go to providers that don't train on prompts (a provider may retain a prompt for a while under its own terms)

## Plans

- Free week: 25 API calls/day for 7 days, no card, up to ~16,000 input tokens per request. After that the key returns 402 upgrade_required until the account has a plan
- Chat: $19/year — the full app (unlimited chat, 100 images/day, 5 agents), no developer API
- Annual: $49/year — 2,000 API calls/day, unlimited chat, 5 agents
- Lifetime: $99 once — 5,000 API calls/day, unlimited chat, 5 agents
- Annual and Lifetime: past the day's calls (or if our quota fails), requests can carry on on the account's own Groq or OpenRouter key, added in Integrations — no limit of ours; responses say X-ZeroLimit-Keys: own
- Every paid plan: 30-day money-back guarantee

## Docs

- [API reference](https://www.zerolimitai.com/docs/api)
- [OpenAPI 3.0 spec](https://www.zerolimitai.com/api/v1/openapi.json)
- [Developer landing](https://www.zerolimitai.com/developers)
- [Pricing](https://www.zerolimitai.com/pricing)
- [Live model ranking](https://www.zerolimitai.com/leaderboard)
- [Live status](https://www.zerolimitai.com/status)
- [Terms and fair use](https://www.zerolimitai.com/terms)
- [Free models list (GitHub)](https://github.com/ClawLabsAI/free-ai-models)

## Contact

- hello@zerolimitai.com