Mobile Chat

Claude Ops exposes an OpenAI-compatible chat endpoint so you can use any OpenAI-compatible app — Open WebUI, ChatBox, LM Studio, or a plain curl — as a mobile or desktop interface to your infrastructure agent.

How it works

When you send a message, Claude Ops:

1. Authenticates your request via Bearer token
2. Extracts the last user message as a prompt
3. Triggers an ad-hoc monitoring session with that prompt
4. Streams the agent's response back in real time (or waits for completion if stream: false)

Stateless sessions

Each request is a fresh session. Claude Ops does not inject your conversation history as context — only the last user message is sent to the agent. This keeps sessions isolated and predictable.

Setup

1. Set the API key

Add CLAUDEOPS_CHAT_API_KEY to your .env file. This is the Bearer token your chat app will use:

CLAUDEOPS_CHAT_API_KEY=your-secret-key-here

If this variable is unset, the endpoint returns 503 (disabled). Pick a strong random value:

openssl rand -hex 32

2. Restart Claude Ops

docker compose restart claudeops

The key is read on every request, so you can rotate it without a restart if you prefer.

Connecting a client

The chat endpoint follows the OpenAI API convention. Point your client at:

Setting	Value
Base URL	http://your-claudeops-host:8080/v1
API key	your CLAUDEOPS_CHAT_API_KEY value
Model	claude-ops (returned by /v1/models)

Open WebUI

1. Go to Settings → Connections → OpenAI API
2. Set API Base URL to http://your-claudeops-host:8080/v1
3. Set API Key to your CLAUDEOPS_CHAT_API_KEY
4. Save — Open WebUI will fetch /v1/models automatically and show claude-ops

curl

# Streaming
curl -N http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer your-secret-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-ops",
    "stream": true,
    "messages": [{"role": "user", "content": "What services are currently unhealthy?"}]
  }'

# Synchronous (wait for full response)
curl http://localhost:8080/v1/chat/completions \
  -H "Authorization: Bearer your-secret-key-here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-ops",
    "stream": false,
    "messages": [{"role": "user", "content": "Restart the nginx container on ie01"}]
  }'

Endpoints

POST /v1/chat/completions

Triggers an ad-hoc session and returns the agent's response.

Authentication: Bearer token (Authorization: Bearer <key>)

Request body:

{
  "model": "claude-ops",
  "stream": true,
  "messages": [
    {"role": "user", "content": "Check the health of all services"}
  ]
}

stream: true — Server-Sent Events (SSE), token-by-token as the agent runs stream: false — Waits for the session to complete and returns the full response

Streaming response (stream: true):

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","model":"claude-ops","choices":[{"index":0,"delta":{"role":"assistant"}}]}

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","model":"claude-ops","choices":[{"index":0,"delta":{"content":"Checking all services..."}}]}

data: {"id":"chatcmpl-...","object":"chat.completion.chunk","model":"claude-ops","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

data: [DONE]

Synchronous response (stream: false):

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "claude-ops",
  "choices": [{
    "index": 0,
    "message": {"role": "assistant", "content": "All services are healthy..."},
    "finish_reason": "stop"
  }],
  "usage": {"prompt_tokens": 0, "completion_tokens": 0, "total_tokens": 0}
}

GET /v1/models

Returns the available model. This endpoint is unauthenticated — clients probe it before asking for credentials.

{
  "object": "list",
  "data": [{"id": "claude-ops", "object": "model", "owned_by": "claude-ops"}]
}

Error responses

Status	Code	Cause
401	invalid_api_key	Missing or wrong Bearer token
400	invalid_request	Malformed JSON or no user message
429	rate_limit_exceeded	A session is already running — try again shortly
503	chat_endpoint_disabled	CLAUDEOPS_CHAT_API_KEY is not set

All errors follow the OpenAI error format:

{
  "error": {
    "message": "A session is already running. Try again shortly.",
    "type": "rate_limit_error",
    "code": "rate_limit_exceeded"
  }
}

Security

• Keep CLAUDEOPS_CHAT_API_KEY out of version control — use .env (which is gitignored)
• Expose the port only on trusted networks; put Claude Ops behind a reverse proxy (Caddy, Nginx) with TLS if you access it over the internet
• Each request runs a full Claude agent session with the permissions of your configured tier — treat the API key like a root credential