nk3750

nk3750 / clawlens

Public

Agent observability and guardrails for OpenClaw — risk scoring, audit trails, dashboard.

clawhub.aiplugins@nk3750openclaw-clawlens agent-observability ai-agents guardrails local-first openclaw
19
1
85% credibility
Found May 21, 2026 at 21 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

ClawLens is an observability dashboard for AI agents. It watches what your agents do, scores the risk of each action, and lets you create rules to block, pause, or monitor dangerous behavior. Everything runs locally on your machine by default, creating a private audit log you can review anytime. The dashboard shows your agents' status, live activity, risk trends, and an attention inbox for items that need review. You can create guardrails directly from observed actions, scoping them to specific agents or your entire fleet.

How It Works

1
🔍 You discover ClawLens

You're running AI agents and want to see what they're actually doing behind the scenes.

2
📦 You install the plugin

One simple command adds ClawLens to your OpenClaw setup, and a local dashboard becomes available.

3
🖥️ You open the dashboard

A clean web interface shows all your agents at a glance, with live status indicators and risk summaries.

4
👀 You watch activity unfold

A live feed shows every action your agents take in real-time, with color-coded risk scores that tell you at a glance which actions are safe and which need attention.

5
You decide how to respond
🚫
Block dangerous actions

Create a rule that automatically stops risky commands before they run.

✅
Require approval first

Set up a rule that pauses dangerous actions and notifies you so you can decide.

🔔
Just get notified

Let risky actions run but receive an alert so you're always informed.

6
📊 You review the audit trail

Everything is logged locally with timestamps, so you can trace exactly what happened and when.

🎉 You're in control

Your AI agents are now visible, their behavior is understandable, and you have guardrails in place to keep things safe.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 21 to 19 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is ClawLens?

ClawLens is an observability plugin for OpenClaw that gives you visibility into what your AI agents are actually doing. It watches every tool call your agents make, scores the risk level of each action, and maintains a tamper-evident audit log. The dashboard shows live sessions, risk breakdowns by agent, and an "Attention Inbox" for anything that needs human review. You can create guardrails directly from observed actions -- block dangerous commands, require approval for risky ones, or just log sensitive operations. Everything runs locally by default, with optional LLM-powered risk evaluation when you want deeper context.

Why is it gaining traction?

The killer feature is creating guardrails from real agent behavior rather than guessing what might go wrong. You watch an agent attempt something risky, click "Add guardrail," and it blocks that pattern going forward. The dashboard is genuinely useful -- not just logs, but a fleet-wide view with risk tiers, session timelines, and activity breakdowns. The audit log uses hash chaining to detect tampering, which matters when you're reviewing incidents. Optional LLM evaluation adds context-aware risk scoring without sending raw data out, and credential redaction is automatic.

Who should use this?

DevOps teams running autonomous agents that execute shell commands, edit files, or call external APIs. Security teams that need audit trails for compliance. Anyone operating multiple agents and wanting operator-controlled guardrails without rewriting agent logic. If you're running agents in production or letting them touch sensitive infrastructure, this gives you the oversight layer that most setups currently lack.

Verdict

ClawLens fills a real gap for agent observability, and the guardrail-from-observation workflow is clever. At 19 stars it's early-stage -- the docs are solid but test coverage and production hardening need real-world mileage. The 0.85 credibility score reflects a small but thoughtful codebase. Worth evaluating if you're already on OpenClaw; for other agent frameworks, watch for broader compatibility or consider this a model for building similar tooling.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.