OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Security & Sandboxing

Dashclaw

Dashclaw is a TypeScript developer tool that governs unattended AI-agent actions through supported enforcement integrations, remote approvals, policy checks and signed execution records.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Dashclaw — OpenViST project profile in Agent Security & Sandboxing.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Dashclaw when an unattended agent needs bounded authority and a review path for higher-risk actions. Select an enforcement integration when the action must be cancellable before execution; cooperative SDK, API and MCP connections require the caller to honor Dashclaw’s verdict.

Who it is for

  • Developers operating AI agents without continuous keyboard supervision.
  • Teams that need human approval before selected destructive or expensive actions.
  • Builders connecting supported agent runtimes to policy-based execution checks.

Documented uses

  • Intercept selected agent actions before execution.
  • Request and resolve remote human approvals.
  • Apply policy checks and standing authorization rules.
  • Record authorization decisions and signed evidence for later review.
  • Inspect enforcement-path liveness and runtime diagnostics.
Before you choose

Requirements & limits

The repository documents MIT-licensed self-hosting and a CLI setup that can be initiated without an account. It identifies CLI, cloud and self-hosted platforms; provider plans and limits may apply to hosting and usage.

  • Fail-closed interception is documented for Claude Code, Codex and Hermes hooks, the OpenClaw gateway and dashclaw_invoke.
  • Bare SDK, API and ordinary MCP integrations are cooperative and depend on the caller honoring the verdict.
  • Approval decisions can be handled through the inbox, CLI, phone PWA, Telegram or Discord.
  • Ed25519 receipts and signed exports provide verifiable record contents, but do not prove that an external effect occurred.
  • The liveness probe reports enforcement-path evidence and runtime diagnostics rather than attestation.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Start a local Dashclaw instance

Initiate the documented CLI setup for a local Dashclaw instance.

Before you start

  • Node 20+ platform marker listed by the repository.
  • A shell that can run the Dashclaw package.
  • Use an interactive shell for the plain command, or the documented non-interactive flags.

Steps

  1. Run the command below in an interactive shell.

    Documented command
    npx dashclaw up

  2. For an agent-operated non-interactive shell, run the command below.

    Documented command
    npx dashclaw up --yes --no-browser

  3. Use the documented health endpoint and the command below to inspect the setup diagnostics.

    Documented command
    npx dashclaw doctor

Expected outcome

The documented setup path is initiated, with the instance API key written to `~/.dashclaw/instance.json`.

Guide limitations

  • This recipe covers setup and diagnostics, not connecting an agent or proving that an external action was executed.
  • The repository warns that the hosted trial should not be attempted headlessly because its captcha requires a human.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
3 October 2026
Original website title
ucsandman/DashClaw · Official homepage title · retained in its source language
How we review sourcesSuggest a correction