AI API for Developers
Build apps, SDK integrations, coding agents, and product workflows with one multi-model AI API.
Last verified: 2026-06-05
Program summary
Who this page is for
AI API for Developers is for application developers, backend teams, and coding-agent users. Use this page to map a real workflow to the correct MetisRouter base URL, endpoint, model ID, pricing unit, and verification path before production volume increases.
- Primary workflows: SDK integration, repo analysis, code editing, RAG, embeddings, and product AI features.
- Recommended model families: frontier reasoning models, coder models, long-context models, embeddings, and low-cost text models.
- Operational proof should come from Usage Logs: request id, model id, endpoint, latency, tokens or assets, and charged USD amount.
Production setup path
A reliable MetisRouter rollout starts with one bounded request, not a full migration. Configure a dedicated API key, choose one published model ID, set output limits for text or agent tasks, and compare the expected pricing unit with the actual usage record after the first call.
- Use https://api.metisrouter.com/v1 for OpenAI-compatible SDKs and tools.
- Use direct provider paths only when the model page documents that endpoint support.
- For high-volume usage planning, email hello@metisrouter.com with the workflow, expected monthly requests, and candidate model IDs.
What to verify
Before scaling a workflow, verify model compatibility, endpoint support, streaming behavior, tool-calling needs, pricing branch, reserve behavior, and support path. The model catalog and Usage Logs should agree on the model ID, endpoint family, usage unit, and billed amount.
- Use exact API model IDs from /models or /docs/models, not display-only names.
- Check image, video, audio, embedding, and token models with their matching endpoints.
- Keep failed request metadata so support can identify balance, authentication, model, upstream, or parameter errors.
FAQ
How should I start with AI API for Developers?
Start with one API key, one exact model ID, one bounded request, and one Usage Logs verification before moving higher-volume traffic.
Where do I find exact model IDs and prices?
Use /models and /docs/models for exact model IDs, endpoint support, pricing summaries, and copyable request examples.