Documentation
An OpenAI-compatible API. If your tool already speaks to OpenAI, you only change two things: the base URL and the key.
- Start in five minutesYour first call, with curl or from your own tool.
- Responses API and Anthropic SDKThe same models through OpenAI's Responses API and Anthropic's Messages API: change the base URL, keep the SDK.
- Guide per tooln8n, Make, WordPress, Python and JavaScript, step by step.
- n8n templatesTen workflows to import: lead scoring, invoices, keywords, meeting notes, brand reports, alerts and a cost sheet.
- Enroutia in your chat appLibreChat, Jan and Open WebUI, with the preset and the three fields they ask for.
- MCPLet your agent create the connection for you
- Error referenceWhat each code means and what to do about it.
- LimitsRequests per minute, maximum sizes, and what happens when you exceed them.
- Catalogue changesWhich models arrive, which leave, and when prices change.
- Credits and balancePacks, the one-time 5 % fee, what each call debits, and what happens at zero.
- AliasesName a model once in your automation; change what it points at from the panel.
- BudgetsThe five ceilings — workspace, project, end client by tag, key and comparison — the codes they answer with, and how a comparison stops at its budget.
- Brand reportsWhat the AI models say about a brand: one POST, one HTML report, one signed webhook.
- Tags and cost per automationTag each call with its workflow; read the bill per automation, export it, and hear when one goes quiet.
- SDKs for Python and JavaScriptThe OpenAI SDK with tags on every call, the router's reason after each one, and a live tail of your project's calls.
- AuditsSend real tasks, approve, replay on cheaper models, and read an anonymised verdict — from the panel or over HTTP.
- Teams and rolesOne project, several people: owner, admin and member, invitations by email, and a selector for those in more than one.
- Privacy and aggregated metricsWhat is never stored, what the automatic checks keep, and what you can choose to share.