Ratchet

io.github.iversonwuweiPublished by GitHub account iversonwuweiv0.12.1Updated Oct 24, 2026

Read-only tools to inspect your agent setup: list MCP servers, flag unpinned ones, test a policy.

VerifiedSTDIODesktop onlyDeveloper ToolsSecurity & MonitoringAI & ML

Overview

AI-generated overview

Ratchet inspects your local agent setup, lists MCP servers, flags unpinned ones, and drafts a least-privilege policy from observed tool use.

What you can do
It runs locally and reads harness configuration to list installed MCP servers and flag those without pinned versions, without executing anything. It can optionally connect to servers to enumerate their tools, ingest real call records to see which tools are never used, and compile a three-state policy (allow, approve, deny) with a stated rationale per entry. It also builds a delivery directory containing the policy, a report, and per-file sha256 manifests, plus a standalone verifier that recipients can run without installing Ratchet. As an MCP server it exposes ratchet_scan, ratchet_policy and ratchet_check over stdio.
When to use it
Use it when you want visibility into which MCP servers and tools an agent actually has, or when you need to turn observed agent behavior into a reviewable least-privilege policy. It suits audits where a recipient must verify results independently rather than trust the sender's logs.
Requirements
A local process on macOS or Linux (amd64 or arm64), installed via install script, Homebrew, container image, or a Claude Desktop one-click package; no Node or Python runtime is needed for the packaged binaries. No accounts, API keys, or environment variables are declared. The optional introspection mode connects to configured MCP servers, and the verifier script needs Python 3.
Before you install
Scanning is read-only, but introspection connects to configured MCP servers and observation collects real call records, which may contain sensitive tool arguments; keep those artifacts local. Capability judgments are name and description heuristics that err in both directions, so review the policy before relying on it. It is not a sandbox and does not block calls at runtime. Feedback output is prepared for you to copy; no telemetry is sent automatically.

Installation

In SourceWeft

  1. Open Ratchet in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

Ratchet

从 agent 的真实行为编译最小权限策略，并给出收货方能自己验证的证据。

bash
curl -fsSL https://github.com/suhui-organization/ratchet/releases/latest/download/install.sh | sh

macOS / Linux 上也可以用 Homebrew（formula 钉住每个平台的 sha256）：

bash
brew install suhui-organization/tap/ratchet

Claude Desktop 用户可以直接装本地 server（一键安装包，不需要命令行）：

下载 ratchet-0.12.0.mcpb 并打开它，Claude Desktop 会完成安装。包内是四个平台的原生二进制 + 一份 manifest.json （MCPB 规范），不需要 Node 或 Python 运行时。

Linux / macOS（amd64 · arm64）。装完先跑这一条——只读配置，不执行任何东西：

bash
ratchet scan --home ~

它会在十秒内告诉你：这台机器上有几个 MCP server，其中几个没锁版本。 （我自己的机器：16 个里 12 个没锁。）

棘轮只能往一个方向转：权限可以收紧，放宽必须显式。

这是什么

一个跑在你自己机器上的工具：它读 agent 用到的工具清单， 编译出一份最小权限策略，再把这次审计的产物打成一份 收货方不需要信任你、也不需要装任何东西就能验证的交付目录。

这不是什么

  • 不是沙箱、不是容器运行时；
  • 不是 prompt 内容过滤器；
  • 不是企业控制台（不做 dashboard、不做多租户、不做 SaaS 依赖）；
  • 不要求你把日志传到哪里去。

为什么值得存在

市场上的产品要么让你自己手写策略（Cedar / CEL / 自然语言）， 要么要你相信它的日志。Ratchet 做的是这两件之间缺失的一步： 策略由观测编译，证据由收货方自己算哈希。

一个例子

bash
# 1) 编译：从工具清单得到一份保守的最小权限策略ratchet policy draft --from examples/inventory.json --out policy.json
# 2) 出交付物：报告 + 逐文件 sha256 清单ratchet-report build --dir ./delivery --policy policy.json
# 3) 收货方验证（不需要装 ratchet）python3 verify.py ./delivery

仓库结构

目录语言是什么
cmd/ internal/Go机器上的引擎：工具清单 → 能力判定 → 最小权限策略。编译成单个静态二进制，目标机器不需要任何运行时
service/Python交付物侧：合规映射、sha256 清单、以及只用标准库的独立验证器
web/Nuxt 4 + Vue 3 + Tailwind公开页面：落地页 + 收货方在浏览器里自己算哈希的验证页
docs/—设计与计划（DESIGN.md / PLAN.md / DECISIONS.md）

状态

0.12.0 — developer preview，链路已经整条打通（本机实测：16 个 MCP server、 12 个未锁版本、164 个工具、策略 allow 43 / approve 106 / deny 15）：

命令做什么
ratchet scan [--introspect] [--share page.html]读 harness 配置列出 MCP server（只读，不执行）；加 --introspect 才连上去列工具
ratchet ingest / observe采集真实调用（hook 出错静默退出），并对照清单看出哪些工具从没被用过
ratchet policy draft / check编译三态策略（每条带判定依据），并试跑一次假设调用
ratchet deliver / report build一条命令出交付目录：策略 + 报告 + sha256 清单
ratchet mcp以 stdio MCP server 形态跑（ratchet_scan / ratchet_policy / ratchet_check）
ratchet feedback把去标识化结果整理成 issue 正文供你复制；没有遥测，不会自动发

拿到它的方式：curl -fsSL https://github.com/suhui-organization/ratchet/releases/latest/download/install.sh | sh、 brew install suhui-organization/tap/ratchet、容器 ghcr.io/suhui-organization/ratchet， 或官方 MCP Registry 里的 io.github.iversonwuwei/ratchet。

清楚的边界：能力判定是名称/描述的启发式，会双向出错（仓里有两条已知误判的测试）； 它不是沙箱、运行时不拦调用；判定不了的会单独列出来请人确认，而不是猜一个结论。 详见 docs/PLAN.md 与 docs/DELIVERY-0.14.0.md。

许可证

Apache-2.0，见 LICENSE。

开发

bash
make test          # 三端全部测试make go-testmake py-testmake web-test

Source: README.md at commit 0b546b9

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

3
  1. v0.12.1LatestSep 20, 2026
  2. v0.12.0Sep 19, 2026
  3. v0.11.0Sep 19, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts