Hermes Agent

Checked against Hermes Agent's docs on 2026-10-11

Hermes Agent connects as a named custom provider: a few lines in ~/.hermes/config.yaml, with the key in ~/.hermes/.env.

Base URL: https://api.creeksidecompute.ai/v1
API: OpenAI Chat Completions
Model id: qwen/qwen3.8-27b, or qwen/qwen3.6-35b-a3b for speed, or meta/muse-glimmer-30b, which always reasons first
Context window: 131072 tokens, up to 65536 of them output
Key: Make one at creeksidecompute.ai/account

Set it up

Make a key. Sign in at creeksidecompute.ai/account with your email and make a key. Each request is paid from prepaid credit: top up by card from $10, or redeem a promotional code.
Store the key. Add this line to ~/.hermes/.env:
~/.hermes/.env
```
CREEKSIDE_API_KEY=<your key>
```

Add the provider to ~/.hermes/config.yaml. If the file's model block has base_url or api_key lines left from another provider, delete them.

~/.hermes/config.yaml

providers:
  creekside:
    api: https://api.creeksidecompute.ai/v1
    key_env: CREEKSIDE_API_KEY
    discover_models: false
    models:
      "qwen/qwen3.8-27b":
        context_length: 131072
      "qwen/qwen3.6-35b-a3b":
        context_length: 131072
      "meta/muse-glimmer-30b":
        context_length: 131072
        supports_vision: true
model:
  default: qwen/qwen3.8-27b
  provider: custom:creekside

The context_length under each model keeps the window when you switch models and back. discover_models: false keeps Hermes's model pickers to these three, and supports_vision: true lets Hermes attach images to Muse Glimmer 30B.

Start Hermes: hermes. To switch to this model inside a session: /model custom:creekside:qwen/qwen3.8-27b. For speed, use /model custom:creekside:qwen/qwen3.6-35b-a3b instead, and for Muse Glimmer 30B, /model custom:creekside:meta/muse-glimmer-30b.

Thinking

Hermes asks for medium thinking unless you say otherwise, so the model thinks by default. To turn it off, add this to ~/.hermes/config.yaml:

~/.hermes/config.yaml

agent:
  reasoning_effort: "none"

In a session, /reasoning none turns it off, and /reasoning low, medium or high turn it on; add --global to keep the choice. /reasoning show and /reasoning hide show or hide the thinking text. Thinking tokens are billed as output.

Muse Glimmer 30B always reasons before it answers, so /reasoning sets how long, not whether: none acts as low, the shortest, and by default it reasons at medium.

What does not work

Image analysis, on the Qwen models. Hermes sends images to other providers it knows. With only a Creekside key it cannot analyse images on them: they are listed for text only.
Side tasks. Session titles and context compression run on this model too, and are billed like any other request.

Troubleshooting

401 invalid API key: The key is wrong, or Hermes did not get it. Check the line in ~/.hermes/.env, then restart Hermes. To test the key alone, curl -s https://api.creeksidecompute.ai/v1/models -H "Authorization: Bearer $CREEKSIDE_API_KEY" returns the model list when the key is good.
404 model … is not served: The model id is wrong. Use qwen/qwen3.8-27b exactly, in lower case, with the qwen/ prefix. GET /v1/models lists every id we serve. With /model, write custom:creekside:qwen/qwen3.8-27b.
400 maximum context length is 131072 tokens: The conversation is longer than the model's window. Keep context_length: 131072 under the model so Hermes compresses the conversation in time, then start a new session.
402 or 429: A 402 means your prepaid credit does not cover the request: add credit at creeksidecompute.ai/account. A 429 means the key's rate, or the number of requests your account may run at once, is used up; agents that work in parallel reach that sooner. Wait and retry. The API reference lists the limits.

Other tools

Sources, read on 2026-10-11: providers, configuration.