Hermes

Hermes is an autonomous AI agent from Nous Research that lives on your server. It speaks the OpenAI API format, so you can point it at Router9 as a custom endpoint — no code changes required.

Setup (recommended: CLI)

Run the interactive model picker from your terminal — not from inside an active Hermes chat session:

hermes model

Select Custom endpoint (self-hosted / VLLM / etc.), then enter:

Hermes writes the endpoint to config.yaml and stores the key in ~/.hermes/.env, so the setting persists across sessions.

To script it instead of answering prompts, set the same keys directly:

hermes config set model.provider custom
hermes config set model.base_url https://api.router9.com/v1
hermes config set model.default anthropic/claude-opus-5   # any id from GET /v1/models

Verify with hermes config show, then start a session:

hermes chat

You can also override the provider and model for a single run:

hermes chat --provider custom --model anthropic/claude-opus-5

Setup (manual: config.yaml)

Hermes keeps its model configuration in ~/.hermes/config.yaml (hermes config path prints the exact location; hermes config edit opens it). Add a custom OpenAI-compatible endpoint under the model: block:

model:
  provider: custom
  default: auto
  base_url: https://api.router9.com/v1
  api_key: sk-r9k-your-key-here
  api_mode: chat_completions
  context_length: 1000000   # optional
  max_tokens: 8192          # optional output cap

When base_url is set, Hermes ignores the provider's built-in endpoint and calls that URL directly. api_mode is chat_completions for Router9's OpenAI-compatible API (Hermes auto-detects it if you omit the key).

Named custom provider

If you point Hermes at more than one endpoint, register Router9 in custom_providers and keep the key in an environment variable instead of the config file:

custom_providers:
  - name: router9
    base_url: https://api.router9.com/v1
    key_env: ROUTER9_API_KEY
    api_mode: chat_completions

export ROUTER9_API_KEY=sk-r9k-your-key-here

Then switch to it mid-session with the /model slash command:

/model custom:router9:anthropic/claude-opus-5

API Key

Router9 authenticates with your sk-r9k- key. Supply it one of three ways:

• hermes model / hermes config set model.api_key ... — Hermes routes secrets to ~/.hermes/.env
• api_key: in the model: block of config.yaml
• key_env: on a named custom provider, with the value exported in your shell or ~/.hermes/.env

Set the key explicitly rather than relying on OPENAI_API_KEY. Hermes may fall back to it for custom endpoints, but that variable also drives Hermes' OpenAI STT/TTS integration — sharing it conflates two unrelated credentials.

Choosing Models

model.default (and --model / /model) take the model id Router9 routes on. Use the exact ids from GET /v1/models — Router9 ids are provider-prefixed slugs, so short names like gpt-4o return 404 model_not_found:

curl https://api.router9.com/v1/models \
  -H "Authorization: Bearer sk-r9k-your-key-here"

A typical hosted catalog looks like:

The exact set depends on your deployment — see Models.

Letting Router9 pick

If your deployment has an auto route configured, set the model to auto and Router9 selects a model for you:

hermes config set model.default auto

auto is not a built-in alias. When no auto route exists it returns 404 model_not_found — check GET /v1/models first.

MCP Setup

Hermes can use Router9's Skills as tools via Router9's hosted MCP server. Add it with the CLI — Hermes prompts for the Authorization header value, where you enter Bearer sk-r9k-your-key-here:

hermes mcp add router9 --url https://api.router9.com/v1/mcp --auth header

Check the connection with hermes mcp list and hermes mcp test router9. See Install MCP Server.

Skills (optional): Router9 also offers image and audio Skills as agent tools. MCP is the recommended way to expose them where your harness supports it; otherwise call them directly as REST endpoints (/v1/image/*, /v1/audio/*, /v1/search/) with your sk-r9k- key — see the API Reference.

Tips

• Router9's flat monthly pricing pairs well with long-running agent sessions — no per-token cost surprises.
• Hermes can expose itself as an OpenAI-compatible API server, so Router9 also backs any frontend you connect to Hermes (Open WebUI, LibreChat, and others).
• Keep your key in ~/.hermes/.env (or a key_env variable) rather than committing it to config.yaml.
• hermes config check flags missing or stale model configuration after an upgrade.