openrouter

Troubleshooting

If requests fail, run zero doctor check-connector --env-name OPENROUTER_TOKEN or zero doctor check-connector --url https://openrouter.ai/api/v1/chat/completions --method POST.

How to Use

OpenRouter exposes an OpenAI-compatible API at https://openrouter.ai/api/v1. Authenticate with a Bearer token. The model field accepts dozens of vendor-prefixed slugs (e.g. anthropic/claude-3.5-sonnet, meta-llama/llama-3.1-70b-instruct, google/gemini-1.5-pro).

1. Basic chat completion

Write to /tmp/openrouter_request.json:

{
  "model": "openai/gpt-4o-mini",
  "messages": [{"role": "user", "content": "Summarize this in one line: VM0 is the agent platform"}]
}

Then run:

curl -s -X POST "https://openrouter.ai/api/v1/chat/completions" \
  -H "Authorization: Bearer $OPENROUTER_TOKEN" \
  -H "Content-Type: application/json" \
  -d @/tmp/openrouter_request.json

2. Try a Claude model

Write to /tmp/openrouter_request.json:

{
  "model": "anthropic/claude-3.5-sonnet",
  "messages": [{"role": "user", "content": "Write a haiku about debugging"}]
}

Then run:

curl -s -X POST "https://openrouter.ai/api/v1/chat/completions" \
  -H "Authorization: Bearer $OPENROUTER_TOKEN" \
  -H "Content-Type: application/json" \
  -d @/tmp/openrouter_request.json

3. List available models

curl -s "https://openrouter.ai/api/v1/models" \
  -H "Authorization: Bearer $OPENROUTER_TOKEN"

4. Streaming completion

Write to /tmp/openrouter_request.json:

{
  "model": "openai/gpt-4o-mini",
  "stream": true,
  "messages": [{"role": "user", "content": "Count to 5"}]
}

Then run:

curl -sN -X POST "https://openrouter.ai/api/v1/chat/completions" \
  -H "Authorization: Bearer $OPENROUTER_TOKEN" \
  -H "Content-Type: application/json" \
  -d @/tmp/openrouter_request.json

5. Check credit balance

curl -s "https://openrouter.ai/api/v1/credits" \
  -H "Authorization: Bearer $OPENROUTER_TOKEN"

Guidelines

1. Model slugs are vendor/model — anthropic/claude-3.5-sonnet, meta-llama/llama-3.1-70b-instruct, etc.
2. OpenAI-compatible — request/response shapes match chat/completions; existing OpenAI client libraries work.
3. HTTP-Referer & X-Title headers — optional but recommended; OpenRouter shows them on the leaderboard.
4. Set route: "fallback" — automatically retry on a backup provider if the primary is down.
5. Pricing is per-model — check https://openrouter.ai/models for current rates before high-volume calls.