npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@runestone-labs/gatekeeper-mcp

v0.1.2

Published

MCP server that runs agent tool calls (shell, files, HTTP, …) through Runestone Gatekeeper — policy enforcement, human approval, and audit for any MCP client

Readme

Runestone Gatekeeper — MCP Server

An MCP server that lets any MCP client (Claude Desktop, Claude Code, Cursor, …) run real tool calls — shell, file writes, HTTP, or any Gatekeeper tool — only through Runestone Gatekeeper. Every call is policy-checked, may require human approval, and is audited. The model never touches your systems directly; Gatekeeper sits in front.

MCP client (LLM)  ──tools/call──▶  gatekeeper-mcp  ──policy-gated──▶  Gatekeeper  ──▶  shell / files / HTTP

Requirements

A running Gatekeeper server (see the repo root). This package is just the MCP front door; it does nothing on its own.

Install / configure

It runs over stdio. Add it to your MCP client config, e.g. Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "gatekeeper": {
      "command": "npx",
      "args": ["-y", "@runestone-labs/gatekeeper-mcp"],
      "env": {
        "GATEKEEPER_URL": "http://127.0.0.1:3847",
        "GATEKEEPER_ROLE": "openclaw"
      }
    }
  }
}

| Env | Required | Default | Meaning | |-----|----------|---------|---------| | GATEKEEPER_ROLE | yes | — | The policy role this server presents to Gatekeeper for every call. | | GATEKEEPER_URL | no | http://127.0.0.1:3847 | Gatekeeper base URL. | | GATEKEEPER_AGENT_NAME | no | mcp-client | Name recorded in audit logs. |

Shared approvals for unattended agents

Local Gatekeeper and this MCP integration are free. Need a teammate to handle approvals while your agents run unattended? We're testing Gatekeeper Team at $49/month, for up to five members and five instances. The proposed Cloud service would coordinate a held action, a teammate's approval or denial, and a record of the outcome. Cloud is planned, not publicly available, and no payment is collected during this test. Only actions actually routed through Gatekeeper are covered; installing the MCP server does not cover a client's other tools.

See the proposed workflow and share your team's needs. The form also lets you optionally report a first real routed action or return use on a later day. These are manual self-reports, not verified external activation. Don't send commands or raw audit logs; this test adds no automatic usage reporting to the MCP server.

Tools

| Tool | Gatekeeper tool | Use | |------|-----------------|-----| | shell_exec | shell.exec | Run a shell command. | | files_write | files.write | Write a file. | | http_request | http.request | HTTP GET/POST (egress-policed). | | gatekeeper_call | (any) | Generic escape hatch — call any Gatekeeper tool (e.g. memory.query) by name. | | gatekeeper_health | — | Check the Gatekeeper server is reachable. |

Every gated call returns one of:

  • allow → the tool output.
  • deny → an error result explaining why (the call did not run).
  • approve → an error result saying a human must approve it in Gatekeeper; the call did not run.

Security model

This server is a policy boundary, so it is deliberately strict:

  • Identity is pinned server-side. Role comes from GATEKEEPER_ROLE; origin is fixed to model_inferred and taint to mcp_client. The MCP client cannot set or override actor / role / origin / taint / capability tokens via tool arguments — there is no code path from a tool argument to any of those.
  • Fail closed. A deny, a pending approval, or any malformed/unexpected Gatekeeper response is surfaced as an error — never as a successful result.
  • No endpoint traversal. Generic tool names are validated as strict dotted identifiers; ../, slashes, and the like are rejected before any request.
  • No leakage. The Gatekeeper URL and anything resembling a credential are scrubbed from error messages before they reach the client.

These invariants are covered by an adversarial test suite (src/*.test.ts).

Distribution

Published to npm as @runestone-labs/gatekeeper-mcp and listed on the MCP Registry as io.github.runestone-labs/gatekeeper (see server.json). The MCP Registry is in preview.

License

Apache-2.0