hermes-provider-requesty

Requesty as a native model provider for Hermes Agent.

hermes plugins install requestyai/hermes-provider-requesty#requesty
hermes plugins enable requesty
echo 'REQUESTY_API_KEY=<your key>' >> ~/.hermes/.env   # https://app.requesty.ai/api-keys
hermes --provider requesty -m claude-sonnet-5-5

Or run hermes model and pick Requesty.

What you get

Model picker	Requesty's managed policies (claude-sonnet-5-5, gpt-6.1-sol, gemini-3.8-flash, …): curated routes with automatic failover. Add @eu to stay on EU-hosted providers. Any model id (anthropic/claude-sonnet-5-5) or org policy (policy/my-policy) also works with -m.
Reasoning	/reasoning and agent.reasoning_effort map onto Requesty's reasoning_effort, sent only to models the catalog marks as reasoning-capable. Defaults to medium.
Prompt caching	Done by Requesty server-side. Nothing to configure.
Tracing	Each Hermes session is sent as requesty.trace_id, so a conversation groups together in the Requesty dashboard.
Context and cost	Context windows and per-token prices (tiered and cache rates included) come from the Requesty catalog, so compression and /usage are accurate.

Configuration

Variable
REQUESTY_API_KEY	Required.
REQUESTY_BASE_URL	Optional. https://router.eu.requesty.ai/v1 for the EU endpoint, or a proxy.

Put them in ~/.hermes/.env. Alias: --provider rq.

Data disclosure

Prompts, tool calls and your REQUESTY_API_KEY are sent to router.requesty.ai (or REQUESTY_BASE_URL), along with the Hermes session id as a trace id. The plugin also reads Requesty's public model catalogs without credentials. It sends no attribution headers and no telemetry.

Development

python -m unittest discover -s tests   # offline; stubs the Hermes runtime
ruff check requesty tests && ruff format --check requesty tests
hermes plugins validate ./requesty

License

MIT