Skip to content

OpenRouter

Autobot supports OpenRouter as a gateway provider. OpenRouter aggregates hundreds of models from multiple providers (Anthropic, OpenAI, Google, Meta, Mistral, and more) behind a single API key and unified billing.

Setup

1. Get an API key

Create an API key at openrouter.ai/keys.

2. Configure credentials

Add your API key to the .env file:

OPENROUTER_API_KEY=sk-or-...

Or use the interactive setup:

autobot setup
# Select "OpenRouter" as provider

3. Configure the provider

In config.yml:

agents:
  defaults:
    model: "openrouter/anthropic/claude-sonnet-4-5"

providers:
  openrouter:
    api_key: "${OPENROUTER_API_KEY}"

4. Verify

autobot doctor
# Should show: ✓ LLM provider configured (openrouter)

Model naming

Models use the openrouter/ prefix followed by the OpenRouter model path:

# Anthropic via OpenRouter
model: "openrouter/anthropic/claude-sonnet-4-5"
model: "openrouter/anthropic/claude-3.5-haiku"

# OpenAI via OpenRouter
model: "openrouter/openai/gpt-4o"
model: "openrouter/openai/o3-mini"

# Google via OpenRouter
model: "openrouter/google/gemini-2.5-pro"

# Meta via OpenRouter
model: "openrouter/meta-llama/llama-3.3-70b-instruct"

# Auto-routing (picks the best model automatically)
model: "openrouter/auto"

The openrouter/ prefix tells autobot to route through OpenRouter's API. OpenRouter model IDs include the original provider as part of the path (e.g. anthropic/claude-sonnet-4-5).

See the full model list at openrouter.ai/models.

Auto-detection

OpenRouter is auto-detected in two ways:

  • By API key prefix — Keys starting with sk-or- are recognized as OpenRouter
  • By API base URL — If api_base contains openrouter, the gateway is detected automatically

This means you can also use OpenRouter without explicit openrouter/ model prefixes if the API key or base URL identifies it.

Configuration reference

Field Required Default Description
api_key Yes — OpenRouter API key (sk-or-...)
api_base No https://openrouter.ai/api/v1/chat/completions Custom API endpoint
extra_headers No — Additional HTTP headers for every request

How it works

OpenRouter acts as a gateway — it accepts OpenAI-compatible requests and forwards them to the upstream provider:

  • Authorization: Bearer header for authentication
  • Standard OpenAI-compatible format for all requests
  • Model prefix — autobot automatically prepends openrouter/ to model names for routing

Since OpenRouter forwards to the original provider, the underlying model's capabilities apply (tool use, context window, etc.). Autobot always uses the OpenAI-compatible path through OpenRouter, even for Anthropic models.

Voice transcription

OpenRouter does not provide a transcription API. If you need voice message support, configure an additional Groq or OpenAI provider for Whisper-based transcription.

Known limitations

  • No streaming — Responses are returned in full after the model finishes generating.
  • No native Anthropic path — Even when using Claude models through OpenRouter, autobot uses the OpenAI-compatible format (OpenRouter handles the conversion).
  • Tool choice is always auto — There is no configuration to force a specific tool or disable tool use per-request.
  • Per-model limits vary — Rate limits and pricing depend on the underlying model. Check openrouter.ai/models for details.

Troubleshooting

Enable debug logging to see request/response details:

LOG_LEVEL=DEBUG autobot agent -m "Hello"

Look for:

  • POST https://openrouter.ai/api/v1/chat/completions model=... — confirms gateway is active
  • Response 200 (N bytes) — confirms API response
  • HTTP 4xx/5xx: ... — API errors with details

Common issues

"No LLM provider configured" — Check that api_key is set and non-empty in config.yml.

"API error: invalid_api_key" — Invalid or revoked API key. Verify at openrouter.ai/keys.

"API error: insufficient_quota" — Out of credits. Top up at openrouter.ai/credits.

"API error: model_not_found" — Model path is wrong or has been removed. Check the OpenRouter models page for current availability.