Netra Logo
Pricing About

Learn / How to use Netra Runtime with Hermes Agent

How to use Netra Runtime with Hermes Agent

Hermes Agent is Nous Research's open-source autonomous agent: it connects to a model endpoint, executes tasks, and improves over time through persistent memory and learned skills. Every chat turn goes to a configured inference provider, and Hermes accepts any OpenAI-compatible endpoint. That makes a Netra Runtime deployment a natural backend, especially since agentic workloads are exactly the long-context traffic Netra is built to serve.

What you need

  • Hermes Agent installed on your machine.
  • A Netra Runtime API key from your dashboard at app.netraruntime.com. The API base URL is https://api.netraruntime.com/v1 (self-hosted deployments use their own endpoint URL instead).

Option 1: the setup wizard

The fastest path is the interactive model wizard:

hermes model
# Select "Custom endpoint (self-hosted / vLLM / etc.)"
# API base URL:  https://api.netraruntime.com/v1
# API key:       your Netra API key
# Model name:    qwen3.6-35b

Hermes writes the configuration to ~/.hermes/config.yaml and stores the key in ~/.hermes/.env. The same options appear during first-time installation, so you can point a fresh install straight at Netra.

Option 2: edit config.yaml directly

For a single default model, set the model block in ~/.hermes/config.yaml:

model:
  default: qwen3.6-35b
  provider: custom
  base_url: https://api.netraruntime.com/v1
  api_key: YOUR_NETRA_API_KEY

If you run several endpoints (say, a hosted Netra deployment and a local test server), give each one a name under custom_providers and keep the key in an environment variable:

custom_providers:
  - name: netra
    base_url: https://api.netraruntime.com/v1
    key_env: NETRA_API_KEY

Then switch mid-session with /model custom:netra:qwen3.6-35b. If your deployment expects provider-specific request fields, add an extra_body map to the provider entry and Hermes merges it into each chat-completions request. Every option is covered in the Hermes provider documentation.

Verify it works

Start a session and ask something. If the reply comes back, tokens are flowing through your Netra endpoint. Common failures: a base URL missing the /v1 suffix (requests 404), a model name that doesn't match an id listed at /v1/models, or a key_env variable that isn't set in the shell running Hermes.

Why pair an agent with Netra

Agents burn tokens: long system prompts, tool outputs, and growing histories make every turn a long-context request. In Netra's published benchmark on a 32K-in, 16K-out workload, it streamed tokens about 5.5x faster than vLLM and finished full answers roughly 3.4x sooner under load. For an agent that loops many times per task, that latency difference compounds. Read more in LLM inference for AI agents.

Frequently asked questions

Does Hermes Agent work with self-hosted inference endpoints?

Yes. Hermes Agent connects to any OpenAI-compatible endpoint through its custom provider path, which covers Netra Runtime deployments whether hosted, in a VPC, or on-prem.

Where does Hermes Agent store its provider configuration?

In ~/.hermes/config.yaml, with secrets in ~/.hermes/.env. You can edit the YAML directly or re-run the hermes model wizard at any time.

Can I switch models mid-session in Hermes Agent?

Yes. Use the /model command in a session, for example /model custom:netra:qwen3.6-35b, to switch to a named custom provider without restarting.

Related