Hermes Agent

Use Fusion Model with Hermes Agent by Nous Research.

Fusion configuration

What is Hermes Agent?

Hermes Agent↗ is an open-source, terminal-native coding and task agent from Nous Research. It supports persistent memory, reusable skills, multiple execution backends, and messaging integrations.

Fusion connects through Hermes' custom OpenAI-compatible endpoint support.

Setup

Before configuring Hermes, follow Getting started for online self-service accounts (see Enterprise onboarding for enterprise accounts) to create and store your API key. If you use FUMO and have a sales representative, enter their Sales Code at signup or bind it later as described in Console and account.

Recommended: use the interactive model selector

Run:

hermes model

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

API base URL: https://capi.fumolab.ai/api/fusion/v1
API key:      your FUMO API key
Model name:   fusion-code-1.0
API mode:     chat_completions

Find the context length in Model specifications. For a service configured by sales, use the limits provided for that configuration. If the value is unconfirmed and Hermes offers automatic discovery, leave it blank.

Manual configuration

If you prefer to edit the configuration directly, first store the key in ~/.hermes/.env:

FUMO_API_KEY=your-key

Then add a named custom provider to ~/.hermes/config.yaml:

providers:
  fumo:
    api: "https://capi.fumolab.ai/api/fusion/v1"
    key_env: FUMO_API_KEY
    transport: chat_completions

model:
  provider: custom:fumo
  default: fusion-code-1.0

Hermes keeps the secret in .env while storing the model, provider, and Base URL in config.yaml.

Start Hermes

Start the classic CLI or the TUI:

hermes
hermes --tui

Hermes will send model requests to the Fusion Chat Completions endpoint.

Model format

With the named provider above, the default model is:

custom:fumo:fusion-code-1.0

Switch the model ID during a session with /model:

/model custom:fumo:fusion-code-1.0
/model custom:fumo:fusion-max-1.0

Hermes requires a model name in its configuration. The public gateway validates the transmitted ID against your account; after acceptance, the Fusion core treats it as a compatibility label and the service configuration associated with the key controls execution.

Auxiliary tasks

Hermes may use auxiliary models for work such as context compression, session titles, and vision analysis. Fusion accepts standard Chat Completions image input, so vision analysis can use the same provider when Hermes sends image_url content parts.

If auxiliary tasks inherit the main model, they will use the same Fusion provider and key. Confirm the behavior on a small task before enabling a long-running workflow.

Common errors

Provider not found

Run hermes doctor and confirm that the provider is named fumo under providers, while model.provider is custom:fumo.

Authentication error

Confirm that FUMO_API_KEY exists in ~/.hermes/.env and that key_env uses the same name.

Model not working

Confirm that Hermes sends an authorized model ID exactly as provided and that the Base URL ends in /v1. The gateway may reject an unknown or unauthorized ID before it reaches the Fusion core.

Context length error

Check the context length in Model specifications or the service configuration provided by sales, then update it through hermes model.

Resources

• Hermes Agent documentation↗
• Hermes provider documentation↗
• Hermes Agent on GitHub↗
• Fusion Chat Completions API

Configure OpenCode

Supported models: fusion-max-2.0, fusion-max-1.0, fusion-code-1.0. Only Max 2.0 supports low / medium / high / max effort (default medium); the other two models do not support adjustment.