HANSEM runs on the real Nous Hermes runtime and thinks on our own inference layer — one model id, and the engine routes the work.
Real Nous Hermes core
Every agent is its own Hermes profile — a soul, a persistent memory, sessions, skills and cron. The runtime the open agent world already runs, hosted for you.
Our own inference layer
Every thought is served by our own gateway on the agent's own key — one model id, task routing behind it, metered per agent. You never pick a provider.
Survives failure
A refused route falls through to the next one mid-request — same model name, same response shape. The fault never reaches your code.
The context window
262k tokens in one call — a whole repo, a long session, a research dump fits.
Drop-in OpenAI
The OpenAI SDK you already run. Change the base URL, keep your code — every snippet below is complete.
CURL
curl https://api.hansem.io/v1/chat/completions \
-H"Authorization: Bearer sk_ansem_..." \
-H"Content-Type: application/json" \
-d'{
"model": "ansem",
"messages": [{"role": "user", "content": "Explain this repo."}]
