# Model API · Quick start

Model API connects your app or AI tool to models from different providers with one Felo API key. Use it for conversations, writing and coding, or choose image and video models for creative tasks. Usage is paid from your platform balance.

## Quick start

1. [Get an API key](https://openapi.felo.ai/console/api-keys) — Create an API key in the console. Copy the full key and keep it safe.
2. [Top up your balance](https://openapi.felo.ai/console/billing) — Buy USD-denominated service credits from $1 and pay for actual API usage. No subscription required.
3. [View all models](https://openapi.felo.ai/models) — Choose a model that supports Chat Completions from the catalog and replace YOUR_MODEL_ID below.

## Use an AI tool

Configure Codex, Claude Code, DeepSeek Harness, OpenClaw, Cursor, Cline, and other agent clients.

- [Codex](https://openapi.felo.ai/docs/agent-clients/codex.md) — Configure Codex CLI with a Felo custom model provider.
- [Claude Code](https://openapi.felo.ai/docs/agent-clients/claude-code.md) — Configure Claude Code to route Anthropic-compatible model calls through Felo API Platform.
- [OpenClaw](https://openapi.felo.ai/docs/agent-clients/openclaw.md) — Connect OpenClaw to Felo API Platform for assistant and agent workflows.
- [Hermes Agent](https://openapi.felo.ai/docs/agent-clients/hermes-agent.md) — Configure Hermes Agent with Felo API Platform model access.
- [OpenCode](https://openapi.felo.ai/docs/agent-clients/opencode.md) — Connect OpenCode to Felo through the OpenAI-compatible endpoint.
- [DeepSeek Harness](https://openapi.felo.ai/docs/agent-clients/deepseek-harness.md) — Connect DeepSeek Harness to Felo through its native DeepSeek adapter or a custom OpenAI-compatible provider.
- [Cursor](https://openapi.felo.ai/docs/agent-clients/cursor.md) — Use Felo API Platform in Cursor through custom OpenAI-compatible settings where available.
- [Cline](https://openapi.felo.ai/docs/agent-clients/cline.md) — Configure Cline with Felo API Platform model access.
- [Pi](https://openapi.felo.ai/docs/agent-clients/pi.md) — Install the Felo Skill for Pi and optionally add Felo as a Pi model provider.
- [TRAE](https://openapi.felo.ai/docs/agent-clients/trae.md) — Use Felo API Platform from TRAE through compatible model provider settings.
- [Roo Code](https://openapi.felo.ai/docs/agent-clients/roo-code.md) — Configure Roo Code with a Felo-compatible model provider.
- [Kilo Code](https://openapi.felo.ai/docs/agent-clients/kilo-code.md) — Configure Kilo Code with Felo API Platform model access.

[Other AI Tools](https://openapi.felo.ai/docs/agent-clients/other-tools.md) · [FeloAPI CLI](https://openapi.felo.ai/docs/cli.md)

## Connect your own application

Keep your OpenAI or Anthropic SDK. Enter the matching Base URL, your Felo API key and a supported model ID.

### OpenAI SDK · Base URL

```text
https://openapi.felo.ai/api/v1
```

### Anthropic SDK · Base URL

```text
https://openapi.felo.ai/api
```

Anthropic SDKs append /v1/messages. Set the base URL to /api, without a trailing /v1. Use your Felo key as the SDK API key.

### Quick start · Chat Completions

Choose a model that supports Chat Completions from the catalog and replace YOUR_MODEL_ID below.

```bash
export FELO_API_KEY="YOUR_FELO_API_KEY"

curl -sS https://openapi.felo.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $FELO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_MODEL_ID","messages":[{"role":"user","content":"Say hello"}]}'
```

HTTP 200 → `choices[0].message.content`

[Set up manually](https://openapi.felo.ai/docs/getting-started.md) · [Protocols and SDKs](https://openapi.felo.ai/docs/guides/protocols.md)

## Protocol and authentication details

Felo-compatible model and video APIs. The model endpoints include OpenAI-compatible Models, Chat Completions, and Responses APIs, plus the Claude-compatible Messages API. Authenticate model and video requests with a Felo API key using Bearer authentication or the X-API-Key header. Submit video tasks asynchronously, poll until status is succeeded, then download the signed output media URL from outputs[]. The media download endpoint is anonymous: each output URL carries a short-lived HMAC signature (exp + sig) and can be opened directly in a browser without an API key. Chat Completions, Responses, and Messages preserve their respective protocol request, response, streaming, and error shapes. GET/HEAD /api/platform/model-catalog is an anonymous, locale-aware directory for model discovery and Playground forms. Inference still requires an API key. All six inference protocols accept X-Felo-Channel; published channels control routing. Images use persistent async tasks and Decisions uses non-streaming state/questions.