One API for the work typical assistants refuse
OpenAI- and Anthropic-compatible endpoints, a catalog of unrestricted and standard models, and a dashboard for keys, usage and balance. Prompts and replies are never stored.
Unrestricted means fewer refusals, not no rules. Read the Acceptable Use Policy
Three request formats, one catalog.
Every model answers in all three formats, at the same price and with the same limits.
/v1/chat/completionsPOST · OpenAIChat Completions
The format most SDKs and agent frameworks send. Bearer key, streaming, tools and JSON mode.
/v1/messagesPOST · AnthropicMessages
The Anthropic format, for the Anthropic SDKs and Claude Code. Key in the x-api-key header.
/v1/responsesPOST · OpenAIResponses
The format Codex speaks. Same base URL and key as Chat Completions.
Run it from the dashboard.
Everything around the API lives in one place: keys, usage, balance and limits.
API keys
Create a key per app or environment and revoke it at any time. The full key is shown once.
Usage per request
Model, tokens, charge and status for every request, without the prompt or the reply.
Prepaid balance
Top up once or take a monthly plan. Each request is charged once, when it finishes.
Rate limits
Each project has a concurrency limit and a requests-per-minute limit, shared by its keys.
Integrations
Setup guides for coding agents, chat frontends and SDKs, each with the exact base URL to use.
Support
Open a ticket from the dashboard and follow the replies in one thread.
What the platform does, and what it does not.
Modellane is an API for running models. It is not a moderation product and not a training service.
Read the Acceptable Use Policy
What we do
- Serve unrestricted and standard models behind one API
- Answer in OpenAI and Anthropic request formats
- Bill per token from a prepaid balance
- Enforce the Acceptable Use Policy at the account level
What we don't do
- Add a refusal or moderation layer of our own on top of any model
- Offer a policy gateway or a configurable content filter
- Train or fine-tune models on your requests
- Host your own model weights
FAQ
Platform questions
- Do I need a new SDK?
- No. The OpenAI and Anthropic SDKs work as they are. Set the base URL and your Modellane key, then use a Modellane model id.
- What is the difference between unrestricted and standard models?
- Unrestricted models refuse far less often than typical assistants. Standard models keep the behavior of their base model. Both use the same API, formats and billing.
- Do you store my prompts or train on them?
- No. Request content is never stored, so there is nothing to train on. We keep a record without the prompt or completion for 180 days for billing, then delete it.
- Is there a moderation layer between me and the model?
- We add no refusal or moderation layer of our own; you apply your own product policy. The Acceptable Use Policy still applies, and accounts that break it are suspended.
- Can I bring my own model?
- No. Modellane serves the models in its catalog. Hosting your own weights or fine-tuned versions is not offered.
- How do I pay?
- From a prepaid balance. Top up once or subscribe to a monthly plan; every request is charged at the model's per-token rate.
Point your client at Modellane.
Create a key, add balance and keep the code you already have.