API reference¶
The Mango Inference API is OpenAI-compatible and also exposes an Anthropic Messages API-compatible endpoint. This section is the authoritative source for endpoints, parameters, and values; guides link here rather than repeating option lists.
Base URL¶
https://api.mangoboost.io/v1
Anthropic-ecosystem tools usually want the host without the /v1 suffix,
because their clients append the path themselves. See
Claude Code for an example.
Authentication¶
All requests require an API key. The same generated static key is used everywhere; how you send it depends on which endpoint you call:
| Endpoint family | Header |
|---|---|
OpenAI-compatible (/v1/models, /v1/chat/completions) |
Authorization: Bearer <your-api-key> |
Anthropic-compatible (/v1/messages) |
x-api-key: <your-api-key> |
A missing, malformed, or unknown key returns HTTP 401. See
Reliability and error handling
for the error body shapes.
Endpoints¶
| Endpoint | Description |
|---|---|
GET /v1/models |
List the models your key can call. |
POST /v1/chat/completions |
Generate a chat completion (OpenAI-compatible). |
POST /v1/messages |
Generate a message (Anthropic Messages API-compatible). |
POST /v1/messages/count_tokens |
Count tokens for a Messages request (Anthropic-compatible). |
These are the whole public surface today. Any other /v1/... path
(/v1/embeddings, /v1/completions, /v1/responses, /v1/files, and the rest
of the OpenAI and Anthropic surface) returns 404. The one Anthropic-side
exception is POST /v1/messages/count_tokens, which is served alongside
/v1/messages. See OpenAI compatibility
for what that means in practice.
Response headers¶
Every API response carries two identifiers. Quote them when reporting a problem through Platform support. They are how the team finds your request.
| Header | Meaning |
|---|---|
x-request-id |
Identifier for the individual request. |
x-trace-id |
Identifier for the request's trace through the platform. |