Claudia OpenAI-Compatible API Documentation

Live on OpenClaw

Claudia OpenAI-Compatible API

Claudia is an OpenClaw-powered assistant exposed through a private, OpenAI-compatible HTTPS API. Applications can call Claudia using familiar chat-completion request formats while the actual response is generated by the Claudia agent running inside OpenClaw.

OpenAI-style

Supports POST /v1/chat/completions and a short POST /v1 alias.

Backed by OpenClaw

The responder is Claudia as configured in OpenClaw, not a raw model endpoint.

Private by default

Bearer token required. The real token is never published in this documentation.

Base URL

Use this base URL for OpenAI-compatible clients:

https://api.claudia-ai.my.id/v1

Full chat-completion endpoint:

POST https://api.claudia-ai.my.id/v1/chat/completions

Authentication

Every API request must include a bearer token:

Authorization: Bearer YOUR_CLAUDIA_API_TOKEN

Do not expose the real token. Store it in an environment variable, secret manager, or server-side configuration. Never put it in frontend JavaScript or public repositories.

Chat Completions

Method	Path	Description
`POST`	`/v1/chat/completions`	OpenAI-compatible chat completion endpoint.
`GET`	`/health`	Service health check.

Request body

{
  "model": "claudia-openclaw",
  "messages": [
    { "role": "user", "content": "Hello Claudia" }
  ]
}

Supported fields

Field	Status	Notes
`model`	Supported	Accepted for compatibility. Internally routed to Claudia/OpenClaw.
`messages`	Supported	Roles and text content are converted into a Claudia prompt.
`user`	Supported	Used to derive a separate OpenClaw session id.
`stream`	Not supported yet	Requests with `stream: true` return an error.

Short Alias

For simple Bash usage, Claudia also supports a short alias:

POST https://api.claudia-ai.my.id/v1

This accepts the same JSON body as /v1/chat/completions.

cURL Examples

OpenAI-compatible endpoint

export CLAUDIA_API_TOKEN="YOUR_TOKEN_HERE"

curl -X POST https://api.claudia-ai.my.id/v1/chat/completions \
  -H "Authorization: Bearer $CLAUDIA_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claudia-openclaw",
    "messages": [
      {"role": "user", "content": "Summarize this in Indonesian."}
    ]
  }'

Short alias

curl -X POST https://api.claudia-ai.my.id/v1 \
  -H "Authorization: Bearer $CLAUDIA_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"Hello Claudia"}]}'

OpenAI SDK Example

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.CLAUDIA_API_TOKEN,
  baseURL: "https://api.claudia-ai.my.id/v1"
});

const response = await client.chat.completions.create({
  model: "claudia-openclaw",
  messages: [
    { role: "user", content: "Hello Claudia" }
  ]
});

console.log(response.choices[0].message.content);

Python

from openai import OpenAI
import os

client = OpenAI(
    api_key=os.environ["CLAUDIA_API_TOKEN"],
    base_url="https://api.claudia-ai.my.id/v1"
)

response = client.chat.completions.create(
    model="claudia-openclaw",
    messages=[{"role": "user", "content": "Hello Claudia"}],
)

print(response.choices[0].message.content)

Response Format

Responses follow the standard chat completion shape:

{
  "id": "chatcmpl-claudia-...",
  "object": "chat.completion",
  "created": 1777145143,
  "model": "claudia-openclaw",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello. I am Claudia."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 587,
    "completion_tokens": 17,
    "total_tokens": 13916
  }
}

Current Limitations

Streaming responses are not supported yet.
Tool-calling JSON compatibility is not implemented yet.
Vision/audio input is not part of this endpoint yet.
Latency can be higher than a raw model endpoint because requests are routed through OpenClaw and Claudia's agent session.

Operations

The API wrapper runs as a systemd service:

systemctl status claudia-openai-api.service
systemctl restart claudia-openai-api.service

The HTTPS reverse proxy is handled by Caddy:

systemctl status caddy
systemctl reload caddy

Health check:

curl https://api.claudia-ai.my.id/health

Security Notes

Keep the OpenClaw Gateway bound to loopback; do not expose port 18789 directly.
Only expose the HTTPS wrapper domain through Caddy.
Use a strong bearer token and rotate it if it is ever shared accidentally.
Do not put the token in browser/frontend code.
Add rate limits before exposing this endpoint to untrusted users.