Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/homenshum/nodebenchai/openclaw_architecturegit clone --depth 1 https://github.com/HomenShum/NodeBenchAIWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.06397 |
| Opus 5 | $0.00000 | $0.03198 |
| Sonnet 5 | $0.00000 | $0.01279 |
| Haiku 4.5 | $0.00000 | $0.00640 |
Grade A, and why
OPENCLAW_ARCHITECTURE scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 631 lines — stays where its author put it; the contents beside it link to each section on GitHub.
OpenClaw Deep Agent Architecture — Self-Diagnosis Reference
Drop this into any codebase. Claude Code (or any agent) can read it, compare against the existing repo, and self-diagnose what to adopt.
NodeBench Reference Status
This document is the reference pattern for the deep-agent / agentic-system model in this repo.
Use it as the conceptual benchmark for:
- multi-level orchestration
- role-specialized agent systems
- boolean judgment gates
- institutional memory
- self-evolution loops
- persistent operator surfaces
In NodeBench, this is a reference architecture, not a claim that the current implementation fully matches OpenClaw yet.
Read this document alongside the current NodeBench implementation surfaces:
convex/domains/agents/core/coordinatorAgent.tsconvex/domains/agents/orchestrator/toolRouter.tsconvex/domains/agents/orchestrator/queueProtocol.tsconvex/domains/agents/orchestrator/worker.tssrc/features/agents/components/FastAgentPanel/FastAgentPanel.tsxsrc/layouts/AgentPresenceRail.tsxsrc/layouts/CockpitLayout.tsx
Interpretation rule:
docs/agents/OPENCLAW_ARCHITECTURE.md= target deep-agent pattern and design language- current Convex + cockpit code = implementation reality and adoption path
System Overview
OpenClaw is an autonomous multi-agent system that continuously monitors, deliberates, self-improves, and executes — with boolean safety gates, institutional memory, and 6 specialized roles. It runs as a "startup-team-in-Slack" that never sleeps.
Core loop: observe → decide → act → learn → evolve
Key differentiator: The system doesn't just execute workflows. It evaluates whether to act (boolean rubric), who should act (role selection), how well it acted (engagement tracking), and how to improve (self-evolution loop).
1. SOUL Contract (Identity + Operating Principles)
Every agent system needs an immutable identity document. This prevents drift and makes behavior auditable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 631 lines · 0 tokens per session scan A 8767537c0843
OPENCLAW_ARCHITECTURE is an agent published in the GitHub repository HomenShum/NodeBenchAI (14 stars, last pushed 19d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 6,397 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
ai-readiness-reporter
Runs the AgentRC readiness assessment on the current repository and produces a self-contained, static HTML dashboard at reports/index.html. Explains every readiness pillar, the maturity level, and an actionable remediation plan, framed by AgentRC measure → generate → maintain loop. Use when asked to assess, audit…
Custom Agent Foundry
Expert at designing and creating VS Code custom agents with optimal configurations.
Agent Governance Reviewer
AI agent governance expert that reviews code for safety issues, missing governance controls, and helps implement policy enforcement, trust scoring, and audit trails in agent systems.
chat-agents
Build AI-powered chat interfaces with AIChatAgent and useAgentChat. Messages are automatically persisted to SQLite, streams resume on disconnect, and tool calls work across server and client.
callable-methods
Callable methods let clients invoke agent methods over WebSocket using RPC (Remote Procedure Call). Mark methods with @callable() to expose them to external clients like browsers, mobile apps, or other services.
webhooks
Receive webhook events from external services and route them to dedicated agent instances. Each webhook source (repository, customer, device) can have its own agent with isolated state, persistent storage, and real-time client connections.