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

@echostash/cli

v0.1.2

Published

Audit an MCP server's tool surface, scan your code for LLM prompts, and gate regressions in CI. No API key, no server, no model calls.

Readme

@echostash/cli

Audit an MCP server's tool surface. Catch prompt regressions in CI.

An MCP server's tool definitions are prompts: the name, description, and inputSchema are the only things a model sees when deciding which tool to call and how to fill its arguments. Reword a description and tool-selection accuracy moves — but the change ships through a PR looking like a docstring edit. Code gets reviewed; prompts don't.

npx @echostash/cli mcp audit https://your-server.example.com/mcp
npx @echostash/cli mcp audit --command "npx -y @acme/mcp-server"

No API key, no account, no server, no database, no model calls. It reads tools/list — and nothing else, so auditing is side-effect free — then analyzes it deterministically.

  acme-support · 2 tools
  protocol 2026-07-28

  Tool Surface Score  97.5/100     context cost ~126 tokens/request

  ! description-thin (1)
      "refund_order" has a 16-character description
      → Say what it does, what it returns, and when to use something else.
  - no-negative-guidance (1)
      "refund_order" never says when *not* to use it
      → A line like "not for X — use Y instead" is the cheapest disambiguation there is.

What it checks

| Check | Why it matters | |---|---| | Token budget | The whole surface ships on every request. This is what it costs you. | | Confusable pairs | Two tools a model can't tell apart is the #1 cause of wrong-tool calls. | | Schema tightness | Open strings where an enum belongs, unbounded arrays, undescribed params. | | Negative guidance | Descriptions that never say when not to use the tool. | | Name hygiene | Generic verbs, mixed casing, near-duplicate names. | | Annotations | readOnlyHint / destructiveHint, and ttlMs / cacheScope cache hints. |

Gate it in CI

Each run writes .echostash/mcp-baseline.<server>.json. Commit it, then:

npx @echostash/cli mcp audit <target> --check    # exits 1 when the score regresses

| Flag | | |---|---| | --check | compare against the committed baseline; exit 1 on regression | | --threshold <n> | allowed score drop before --check fails (default 0) | | --update-baseline | accept the current state while checking | | --header k=v | extra HTTP header (repeatable) | | --env k=v | variable to hand a --command server (repeatable) | | --inherit-env | forward your entire environment to the server (off by default — see below) | | --from-file <p> | audit a recorded surface instead of connecting | | --json | machine-readable output |

Also included

echostash scan [dir] — find LLM prompts in your own codebase by locating the call sites they flow into. Language-agnostic, no annotations, no SDK. See the main repo.

MIT

Environment of a spawned server

--command servers get a minimal environment (PATH, HOME, and the handful of variables a process needs to start) — not your shell's. An audited server is usually somebody else's code, and it must not inherit OPENAI_API_KEY, NPM_TOKEN, etc. by accident. Pass what the server actually needs with --env, or opt into full inheritance with --inherit-env when you trust it.