# DashClaw — Decision Infrastructure for AI Agents

> DashClaw intercepts agent actions through MCP and SDK paths, exposing governance tools, approvals, and audit-ready decision trails for production use.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

```bash
npm install dashclaw
pip install dashclaw
npx -y @dashclaw/mcp-server --help
```

## Intro

DashClaw intercepts agent actions through MCP and SDK paths, exposing governance tools, approvals, and audit-ready decision trails for production use.

**Best for:** teams that need approvals, policy checks, and traceable governance across MCP hosts and custom agent frameworks

**Works with:** Claude Code, Claude Desktop, managed-agent MCP hosts, Node and Python SDKs, framework-driven agent execution

**Setup time:** 10-18 minutes

### Key facts (verified)

- GitHub: 257 stars · 14 forks · pushed 2026-05-13.
- License: MIT; owner avatar verified from GitHub API for ucsandman.
- Entry point checked from README: `npm install dashclaw`.

## Main

DashClaw is useful when ordinary logging is not enough. It sits on the action path and gives you a chance to inspect or block risky behavior before execution.

The repo is concrete about scope: 8 governance tools, 4 resources, stdio MCP for desktop-style hosts, and SDKs for explicit framework integration.

That makes it a pragmatic bridge between lightweight local agent experiments and regulated production automation, where an approval step needs to be part of the design rather than an afterthought.

### Source-backed notes

- README lists MCP server and SDK integration paths.
- The MCP package exposes 8 governance tools and 4 resources.
- SDK install examples are `npm install dashclaw` and `pip install dashclaw`.

### FAQ

**Q: Does DashClaw require MCP?**
A: No. MCP is one path, but the project also offers Node and Python SDKs.

**Q: What does the MCP package expose?**
A: The README says 8 governance tools and 4 resources.

**Q: Why adopt it?**
A: Because governance is enforced on the action path, not only described in policy documents.

## Source & Thanks

> Source: https://github.com/ucsandman/DashClaw
> License: MIT
> GitHub stars: 257 · forks: 14

---

<!-- ZH -->

## Quick Use

```bash
npm install dashclaw
pip install dashclaw
npx -y @dashclaw/mcp-server --help
```

## Intro

DashClaw 通过 MCP 与 SDK 路径拦截 agent 行为，提供治理工具、审批控制与可审计决策轨迹，适合真实生产环境接入。

**Best for:** 需要在 MCP 客户端或自定义 agent framework 里加入审批、策略检查与可追踪治理能力的团队

**Works with:** Claude Code、Claude Desktop、托管式 MCP host、Node 与 Python SDK、自定义 agent framework

**Setup time:** 10-18 minutes

### Key facts (verified)

- GitHub：257 stars · 14 forks；最近更新 2026-05-13。
- 许可证：MIT；作者头像与仓库链接均已通过 GitHub API 复核。
- README 中核对过的入口命令：`npm install dashclaw`。

## Main

DashClaw 的意义在于它不只是“记日志”，而是站在动作执行路径上，让你在高风险操作发生前就能检查、审批或阻断。

README 对边界写得很明确：8 个治理工具、4 个资源、适用于桌面类 host 的 stdio MCP，以及面向框架集成的 SDK。

这让它成为轻量本地实验与受监管生产自动化之间的现实桥梁，因为审批步骤从设计之初就被纳入体系，而不是事后补丁。

### Source-backed notes

- README lists MCP server and SDK integration paths.
- The MCP package exposes 8 governance tools and 4 resources.
- SDK install examples are `npm install dashclaw` and `pip install dashclaw`.

### FAQ

**问：Does DashClaw require MCP?**
答：No. MCP is one path, but the project also offers Node and Python SDKs.

**问：What does the MCP package expose?**
答：The README says 8 governance tools and 4 resources.

**问：Why adopt it?**
答：Because governance is enforced on the action path, not only described in policy documents.

## Source & Thanks

> Source: https://github.com/ucsandman/DashClaw
> License: MIT
> GitHub stars: 257 · forks: 14


---
Source: https://tokrepo.com/en/workflows/dashclaw-decision-infrastructure-for-ai-agents
Author: MCP Hub