ainfera-hermes — hermes-agent + Ainfera

Hermes integration + Ainfera Routing. Same agent loop — 2 env vars. Routed inference + signed audit underneath.

Two env-var change

export HERMES_INFERENCE_BASE_URL=https://api.ainfera.ai/v1
export HERMES_API_KEY=ai_infera_...                      # https://app.ainfera.ai/signup

Your existing hermes-agent code keeps working. You now have:

• Signed Agent Card identity (L1)
• Per-call budget enforcement on routed inference (L3)
• Hash-chained audit per call, verifiable offline (L4)
• 5 frontier models via one key: Claude Opus 4.7 · GPT-5.5 · Gemini 3.1 Pro · Grok 4 · Mistral Large 3

EU AI Act Annex IV ready — every call hash-chained, signed, exportable.

Quickstart

If you have hermes-agent installed:

git clone https://github.com/ainfera-ai/ainfera-hermes
cd ainfera-hermes
export AINFERA_API_KEY=ai_infera_...
./main.sh

Or with curl only — no hermes-agent install needed:

AINFERA_API_KEY=ai_infera_... ./curl-example.sh

What this proves

Pointing hermes-agent at Ainfera adds signed identity, Ainfera Routing, and a verifiable audit log without changing your agent loop or prompt structure. The HERMES_INFERENCE_BASE_URL hook routes every Inference through Ainfera; Ainfera signs the receipt and writes the AuditEvent into your Agent's chain.

See the Hermes Pool landing page for the full pitch + companion concept pages.

Verify the audit chain

After running, each call returns a receipt URL. You can also verify offline:

pip install ainfera-verify
ainfera-verify chain <your-agent-id>

Disclaimer

hermes-agent is the NousResearch framework. This repository is not affiliated with NousResearch — it's a curated landing for builders who already use it.

License

Apache 2.0. See LICENSE.