Reference documentation. Available models, billing and routing depend on the current site configuration.

懂api / START HERE

Find the task, then run the request

Create a key, configure a client, call an API, or recover a failed request. Each route contains only the settings, test, and success signal needed for that task.

Last updated: 2026-10-17

Make the first call Configure a client

ROUTING ENDPOINTS LIVE

STANDARD / text and streaming http://localhost:3000/v1

DIRECT / long-running generation http://localhost:3000/v1

TASK ROUTES

What do you need to finish?

Each route has a stable URL that can be shared with a teammate.

First call

Quick Start

Create a key, read the live catalog, and verify the first HTTP 200.

Start → Tools

Client setup

Configure CC Switch, Codex CLI, Claude Code, Cursor, and compatible desktop clients.

Choose a client → Code

API reference

Use Chat Completions, Responses, Anthropic Messages, image, and video routes.

View routes → Cost

Pricing and billing

Understand official USD numbers, group multipliers, cache dimensions, and actual charges.

Check pricing → Choose a group

Group selection and questions

Compare GPT and Claude groups by use case, cache, stability and price tier; understand hourly overview and statistics.

Choose a group → Routing

Smart routing

Select a GPT or Claude smart text group and understand failover. Images and Seedance use regular groups.

Configure routing → Generation

AIGC Workshop

Choose currently available image and video models, submit tasks, and retrieve results.

Use the Workshop → Recovery

Troubleshooting

Isolate 401, 404, 429, 5xx, model routing, streaming, and timeout failures.

Find the failure →

ONE KEYUnified authentication

Use the same 懂api key across compatible APIs.

MODEL IDCopy live IDs

Choose models from the live catalog instead of guessing names.

REQUEST LOGVerify with evidence

Confirm HTTP 200 and the matching usage log.