API overview
Choose the Gateway API for model requests or the Console API for organization management.
Edgee exposes two APIs with different credentials and responsibilities.
| API | Use it for | Credential |
|---|---|---|
Gateway API, https://edgee.io | Model requests, streaming, and token operations | Gateway API key |
| Console API | Organizations, Gateway keys, provider keys, and usage exports | Personal API token |
Start with your first Gateway request or Console API setup. A provider credential registered through BYOK does not replace either Edgee credential.
Choose a request format
| Endpoint | Client or task |
|---|---|
POST /v1/chat/completions | OpenAI Chat Completions clients |
POST /v1/responses | OpenAI Responses clients |
POST /v1/messages | Anthropic Messages clients |
| Models | Discover available models |
| Count tokens | Estimate text tokens locally |
| Anthropic token count | Count tokens using Anthropic’s provider tokenizer |
| Compress | Use the compression endpoint |
Choose the format expected by your existing client. Model capabilities and supported request fields vary; API compatibility does not mean every provider implements every feature. Use a canonical author/model identifier from the model catalog where required.
For OpenAI clients, the usual base URL includes /v1: https://edgee.io/v1. Anthropic clients append their own endpoint paths; follow the Anthropic SDK guide.
Verify and diagnose
After sending a request, find it in Logs for the same organization and API key. Inspect the served model/provider and cost. Routing settings can select a different model from the one requested.
For a failed request, inspect both the HTTP status and error body. A 429 caused by a spending limit needs a different response from a transient rate limit. See Authentication, Errors, and Usage limits.