Hermes agent
Hermes agent
Hermes (NousResearch) reads provider
config from ~/.hermes/config.yaml and secrets from ~/.hermes/.env. Add Hyphen
as a custom OpenAI-compatible provider whose base URL ends in /v1 — Hermes
appends /chat/completions and talks to the gateway's default surface.
1. Get a key
app.hyphen-solution.com → sign in →
Dashboard → API Keys → Create. Copy the sk-hy-... key.
2. Configure — one copy-paste block
Add this to ~/.hermes/config.yaml:
Code
Then put the key in ~/.hermes/.env. Hermes resolves credentials by
appending _API_KEY to the provider name — for a provider named hyphen:
Code
That's the full setup. Fields:
base_url— the OpenAI-compatible root, ending in/v1. Hermes appends/chat/completionswhen it calls the model.model— a Hyphen model (m3strongest,m2.7balanced,m2.5fast).context_length— the context window to advertise for this provider/model.
3. Verify
Code
Ask it to reply with a single word and confirm you get a completion. With Hyphen
set as the default provider above, plain hermes chat also works.
Related: hermes proxy
Hermes can run as a local proxy (hermes proxy) that exposes an
OpenAI-compatible endpoint on your machine and forwards to a configured
provider. If you front Hyphen this way, point the proxy's upstream provider at
the same base_url and key above — everything downstream then targets
http://localhost:<port>/v1 while Hyphen remains the real backend.
Troubleshooting
- Auth errors — confirm
HYPHEN_API_KEYis in~/.hermes/.envand matches the provider name (hyphen→HYPHEN_API_KEY). - 404 / wrong path —
base_urlmust end in/v1. 429— monthly budget spent; see Rate limits & caps.- Config keys moved? Hermes keys (
custom_providers,context_length, thehermes proxysubcommand) evolve quickly. Confirm the current shape at the Hermes configuration docs.