Agents

Use this section when your AI tool is a long-running agent — typically a runtime that loads provider config from a YAML/JSON file rather than reading environment variables on every call. WorldRouter exposes an OpenAI-compatible inference endpoint that these agents register as a custom provider, then route traffic through.

This section currently covers Hermes Agent. Additional agent integrations land here as further runtimes adopt WorldRouter.

Shared prerequisites

• Create a WorldRouter API key in the API Keys dashboard.
• Make sure the team has available credits on the Credits page.
• Know your inference base URL — copy it from the Base URL field on the API Keys dashboard. All snippets below assume the same base URL with the /v1 suffix.

Common configuration pattern

Agent runtimes typically store provider configuration in a structured file (YAML, JSON, or TOML) rather than environment variables. WorldRouter integrates the same way regardless of agent:

• Provider name — a label you pick (worldrouter, infer, etc.) — used elsewhere in the agent config to select the provider.
• Base URL — your WorldRouter inference endpoint, including the /v1 suffix.
• API key — your WorldRouter API key.
• API mode — OpenAI-style chat completions.
• Models — the WorldRouter model IDs you want the agent to be able to select from. Get the canonical list from the Models page.

Pick a guide below for the exact config schema your agent expects.