Plain English: what Hermes Loop actually does.Hermes does the reasoning. Hermes Loop runs and proves the work.
Hermes Loop is the operator surface for the Hermes Agent. You describe a job, the system runs a crew of Hermes-driven subagents, gives them governed tools (browser, search, terminal, Python, vision, image, audio, memory), gates risky outputs behind your approval, and produces a hashed receipt that proves what happened.
The control room for the Hermes Agent.
Hermes Agent is Nous Research's open-source autonomous agent — it lives on your server, remembers what it learns, and gets more capable the longer it runs. Hermes Loop is the operator surface: you launch missions, run named crews of subagents, gate risky outputs with approvals, and produce a hashed receipt that proves what happened.
The autonomous engine: persistent memory, subagents with their own conversations and terminals, natural-language cron, and native tools (web search, browser automation, vision, multi-model reasoning).
The governance layer: crews, a job queue, approvals, hashed receipts, a trust ledger, evals, schema self-correction, real-cost accounting, and a full audit trail.
A chatbot gives an answer. The Hermes Agent does the work autonomously; Hermes Loop proves it: who ran, which tools were called, which memory was used, what was approved, and what hash signed off the run.
Run Bug Hunter on /demo-target, approve the QA report, generate the receipt, then open Trust to see the run roll up into risk and proof.
The pieces
The reasoning. Multi-model routing (fast/strong/judge/vision) maps each agent to the right model.
Ordered subagents with system prompts and Zod-validated outputs. Built-in or custom.
Browser (Playwright), terminal_exec, python_rpc, web search (Tavily), vision (Gemini), image generation, ElevenLabs TTS, memory. All sandboxed and policy-gated.
Risky outputs pause until you decide. Drafts/trades/exports never leave on their own.
Per-mission proof: agents, tools, approvals, memory, real cost, integrity hash.
System-wide reliability roll-up. Reads from receipts.
Real Hermes runs end to end. Real assertions on receipts. Exit code gates CI.
POST /api/integrations/webhook → Inbox → Triage → Mission. Optional HMAC signing.
After a mission settles, Hermes Loop distils up to 3 reusable lessons into Skill rows scoped to the crew. Future runs inject them automatically.
Every mission queries operator-approved memory across all prior sessions. Top matches inject into the Triage Agent + first-step prompt.
Native MCP client. Set MCP_SERVERS to a JSON list of remote MCP servers; their tools appear in the agent tool registry.
Hermes Agent is the engine. Hermes Loop is the control room.
Every Hermes Agent capability has a corresponding surface in this app that governs, audits, and proves it. Hermes Agent gives you autonomy; Hermes Loop gives you governance and proof.
What's wired right now
Every row below reads from the actual env. Statuses are live — if a provider isn't set up, the row tells you which env var to add. Hermes Loop never claims a feature is wired when it isn't.
What you can click to test it
Real Playwright crawl of /demo-target with screenshots + accessibility findings.
Read-only diagnostic crew using terminal_exec under policy.
Live Tavily-backed search. Brave + SerpAPI also wired as fallbacks.
Multimodal analyze on Gemini 2.5 Flash via Hermes. Image URL + task → structured JSON.
Image-capable Hermes models (Gemini, Flux). Fal/Replicate as fallback.
ElevenLabs primary. Returns a real MP3 in ~2 seconds.
POST /api/integrations/webhook → Inbox → triage → mission.
Source of truth on shipped vs partial vs missing.
13 stops from health probe to deployment env vars — live status.
What Hermes Loop adds on top of Hermes
- Approvals. Risky outputs (drafts, trades, exports, gated tool calls) wait for you. /approvals
- Workflow receipts. Every mission ends with a hashed timeline and real cost. /receipts
- Trust ledger. Aggregate reliability across runs. /trust
- Memory governance. Operator-approved context with usage records and diffs. /memory
- Evals harness. Real-Hermes test suite — exit code = failed cases.
- Real-cost accounting. Real per-model rates fetched live, recorded on every receipt.
- Learning loop. After every mission settles, Hermes Loop distils up to 3 reusable lessons into /skills. Future runs of the same crew inject them into their system prompt.
- Cross-session memory recall. Every mission queries operator-approved memory across all prior sessions; the top matches are injected into the Triage Agent and first-step prompts.
- MCP integration. Native MCP client. Set
MCP_SERVERSand remote tools show up in the registry alongside built-ins.