+++++++++++

01 >> AI OBSERVABILITY

02 >> AGENT OVERSIGHT

03 >> POLICY ENGINE

CRABWISE.AIv0.0.1-ALPHA

Crabwise

Govern your
agents before
they act

Monitor what your agents are doing, set rules they can't break, and maintain a complete audit trail across providers.

Crabwise-AI/Crabwise

crabwise — zsh

Crabwisev0.1.1-alphaLocal-first AI Agent Governance

─═─═─═─═─═─═─═─═─═─═─═─═─═─═─═─═─═─═─══─══─═

◉Daemonrunning◉Log watcherenabled◉Proxy127.0.0.1:9119◉Uptime26h 15m 8s

crabwise — zsh

Crabwisev0.1.1-alphaLocal-first AI Agent Governance

─═─═─═─═─═─═─═─═─═─═─═─═─═─═─═─═─═─═─══─══─═

◉Daemonrunning◉Log watcherenabled◉Proxy127.0.0.1:9119◉Uptime26h 15m 8s

+++++++++

04 >> ABOUT

Infrastructure
level control

Commandments are governance rules (policies) for AI agents. Block destructive commands, flag credential access, enforce approved models. Customize for your own use-case.

Plain YAML

Hot Reloadable

Easily Configured

Commandment Templates

zsh

~ crabwise service inject

--agent openclaw

✓ intercepting via proxy

openclaw

🦞

POLICIES

GitHub

OpenAI

AWS

zsh

~ crabwise service inject

--agent openclaw

✓ intercepting via proxy

openclaw

🦞

POLICIES

GitHub

OpenAI

AWS

Local-first agent
control plane

Status

PlatformMacOS, Linux

AgentsClaude, Codex, OpenClaw

EnforcementHTTPS proxy

AuditSQLite + hash chain

Crabwalk

Coming Soon

Real-time observation layer. Watch your agents act.

Try the Prototype

Commandments

Policies your agents can't break. Define boundaries once, enforce everywhere.

Audit trail

Immutable, structured logs. Every action timestamped and attributable across sessions.

Multi-provider

One interface across Claude, Codex, OpenClaw + more to come.