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

@archstone/agent

v0.19.1

Published

Embed Archstone capabilities directly in your own agent loop: typed tool definitions for Anthropic, OpenAI, Gemini and plain JSON Schema, plus fail-closed execution over a compiled IR — no MCP server process required.

Readme

@archstone/agent

Embedded agent SDK (RFC-0008 / ADD-0008): load a compiled Archstone IR artifact, generate typed tool definitions (Anthropic · OpenAI · Gemini · JSON-Schema), invoke capabilities with fail-closed response mapping, and judge model-produced documents at the extraction boundary — all without running a separate MCP server process.

Two entry points:

  • Root (@archstone/agent): fromIR(json) constructs an embedded instance; tools(format) generates typed tool defs; execute(capabilityId, input) invokes a capability with mapping applied; extractor(resource, format) binds one declared resource to one target — the schema a model is given and the boundary its answer passes through, as one object (ADR-0011). Zero MCP SDK, zero node:fs — safe to bundle and run anywhere (browser, Worker, native app).
  • /mcp subpath (@archstone/agent/mcp): mcpHandler(archstone, {bearerToken}) — a mountable, fail-closed Streamable-HTTP MCP endpoint for consumers who want to expose an embedded instance as an MCP server (e.g., Claude API mcp_servers, ChatGPT connectors).

Quick start

import { fromIR, tools, execute } from "@archstone/agent";

// Load a compiled IR (produced by `archstone build manifest/`)
const ir = JSON.parse(fs.readFileSync("archstone.ir.json", "utf-8"));
const archstone = fromIR(ir);

// Get tool definitions in your preferred format
const anthropicTools = archstone.tools("anthropic");
const openaiTools = archstone.tools("openai");
const geminiTools = archstone.tools("gemini");

// Invoke a capability — accepts both raw dotted id or sanitized tool name
const result = await archstone.execute("tourism.search", {
  location: "Paris",
  checkInDate: "2026-08-01",
});
// Same call with sanitized tool name (as returned by tools()):
// const result = await archstone.execute("tourism_search", {...});

if (result.status === "ok") {
  console.log("Success:", result.data);
} else if (result.status === "degraded") {
  console.log("Partial:", result.data, "Missing:", result.degraded);
} else if (result.status === "violation") {
  console.log("Contract violation:", result.missing);
} else {
  console.log("Error:", result.error);
}

Extracting business data from unstructured input (ADR-0011):

A Resource Definition already says what a provider's response is mapped into. The same definition says what a model must produce, and extractor() carries both halves — the closed schema the model is given, and the validator its answer is judged by — so the two cannot drift apart.

const stay = archstone.extractor("tourism.Stay", "anthropic");

// Give the model the schema — either axis, same schema underneath.
stay.structuredOutput;   // native structured output, in the envelope your provider expects:
                         //   Anthropic: output_config.format
                         //   OpenAI:    text.format
                         //   Gemini:    response_format

// ...or extraction as a forced tool call, in the same envelopes tools() emits. The
// instruction is required and never defaulted: it says what to DO on this occasion, which
// no Resource Definition can know.
stay.tool("Record the stay described in this booking email.");

// Judge what comes back, at a boundary Archstone owns.
const result = stay.validate(modelOutput);

if (result.status === "ok") {
  save(result.data);
} else if (result.status === "degraded") {
  save(result.data); // declared optional fields absent: result.degraded
} else {
  reject(result.missing, result.invalid); // withheld whole — result.data is absent
}

// Undeclared keys never reach `result.data`, at any depth, and are named in
// `result.undeclared`. There is no passthrough option, for the same reason there is none
// in ADR-0008.

status proves shape, never truth: a model that invents a plausible, correctly-typed value returns ok, and nothing at this boundary can tell that apart from a real one. Error messages never echo a value from the document — the extraction input is by construction the most sensitive text in the deployment.

For HTTP-based MCP:

import { mcpHandler } from "@archstone/agent/mcp";

const handler = mcpHandler(archstone, {
  bearerToken: process.env.ARCHSTONE_TOKEN,
  invoke: { env: process.env }, // optional: env injection for REST connectors
});

// Mount on your framework (e.g., Hono, Express, fetch-based)
const response = await handler(new Request(...));

Part of Archstone, an open-source Capability Platform. For full documentation, see archstone/packages/agent and the main README.md.

License

Apache-2.0