# Agent Runtime

This static explainer maps a standard tool-using agent runtime loop across NousResearch/hermes-agent and openclaw/openclaw.

## Canonical Loop

1. Platform input: normalize an inbound platform event into a stable session identity.
2. Session hydration: assemble the prompt, message history, compaction summaries, and memory context.
3. Agent loop: run the model turn, enforce iteration or stop policy, and decide whether more tool work is needed.
4. Tool dispatch: resolve tool schemas, apply tool policy, execute calls, and return tool results to the loop.
5. Persistence: append the transcript, tool results, summaries, and memory-bearing state for the next turn.

Inbound event and outbound response are the outer rails around this five-part runtime.

## Implementations

- Hermes Agent: Python gateway + AIAgent + tools.registry + SQLite session DB.
- OpenClaw: TypeScript ACP/session layer + CoreAgentHarness + agentLoop + JSONL session tree.

## Hermes Agent Source Anchors

- run_agent.py: AIAgent, persistence hooks, tool-call forwarding.
- agent/conversation_loop.py: max-iteration loop and provider calls.
- agent/tool_executor.py: sequential/concurrent tool execution.
- tools/registry.py: self-registering tool schemas and dispatch.
- gateway/session.py: SessionSource identity.
- gateway/platforms/base.py: platform adapter boundary.

## OpenClaw Source Anchors

- packages/agent-core/src/agent-loop.ts: core LLM/tool loop.
- packages/agent-core/src/harness/agent-harness.ts: session hydration, hooks, persistence.
- packages/agent-core/src/harness/session/session.ts: context assembly from session branch.
- packages/agent-core/src/harness/session/jsonl-storage.ts: append-only JSONL storage.
- packages/acp-core/src/session.ts: ACP session store.
- packages/acp-core/src/runtime/types.ts: runtime session and turn contracts.

## Machine-Readable Map

Agent manifest: ./agent.json
JSON runtime map: ./agent-runtime-map.json