W3 Agent Safety

LOCAL. OFFLINE. ZERO RUNTIME DEPENDENCIES.

Stop the agent before it signs, destroys, or leaks.

A deterministic command guardrail and project risk check for Codex and Claude Code operating near Kubernetes, Terraform, Foundry, and production secrets.

Buy the $1 Risk Check See what it blocks

Pay with Base USDC or request a PayPal invoice from dev.koriel@gmail.com. Product delivery follows payment verification.

policy check 
$ was check --command \
  "cast send 0x71… 'upgradeTo(address)'"

DENY 100 web3.cast.broadcast

$ was check --command "terraform apply tfplan"

ASK 60 infra.terraform.apply

$ was check --command "terraform plan"

ALLOW 0

0 network calls

0 secret values uploaded

106 automated tests

2 agent harnesses

THE FAILURE WINDOW IS ONE COMMAND

Put policy in front of the shell.

Safe inspection continues without noise. Infrastructure mutation asks or blocks according to the harness. Known destructive and signing patterns stop before execution.

01

Command policy

Classifies compound commands, nested shells, wrappers, absolute paths, and quoted arguments.

02

Project audit

Checks Codex, Claude Code, MCP, and secret-like filenames without reading secret contents.

03

Native hooks

Returns structured allow, ask, and deny decisions through current PreToolUse contracts.

DEFAULT POLICY

Opinionated where mistakes are irreversible.

DENY wallet broadcasts and Forge `--broadcast`

DENY Terraform destroy and namespace deletion

DENY private-key reads and pipe-to-shell

ASK infrastructure and Kubernetes mutation

ALLOW plans, reads, tests, and contract calls

48-HOUR LAUNCH PRICING

Start at $1. Deploy the control your team needs.

Four organization launch slots are available. Each paid tier is for internal use under the commercial license.

Risk Check

$1

One project, one local audit, one report.

  • Risk score
  • Config findings
  • Remediation order
Buy Risk Check

Individual

$249

Full guardrail for one operator.

  • Command hook
  • Project audit
  • Field manual
Buy Individual

TEAM DEFAULT

Team

$1,499

Up to ten named operators.

  • Everything in Individual
  • Team policy baseline
  • 60-minute install review
Buy Team

4 LAUNCH SLOTS

Organization

$2,500

One legal entity, internal use.

  • Unlimited internal seats
  • Signed policy baseline
  • Deployment review
Buy Organization

One layer, clearly scoped.

Web3 Agent Safety catches known command patterns. It is not a security boundary. Keep sandboxing, least-privilege credentials, hardware wallets, multisig policy, review, and deployment approval in place.

INSTALL IN MINUTES

Your agent moves fast. The dangerous command should not.

Message @devkoriel