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

dfx-intel

v0.1.1

Published

Evidence-backed private markets research for agents: DFX Intelligence over MCP and A2A. Zero runtime dependencies.

Readme

dfx-intel (TypeScript)

A small client for DFX Intelligence, an evidence-backed research service on private markets: private credit, private equity, venture, family offices, registered investment advisers, public pension allocators and US commercial real estate. It is written for agents. Every answer carries its claims, the URLs of the documents those claims come from, what is uncertain, and what lies outside coverage.

It has no runtime dependencies and uses global fetch, so it runs on Node 18+, Deno, Bun, browsers and edge runtimes.

npm install dfx-intel

Quick start

import { research } from "dfx-intel";

const r = await research("Which BDCs lend to Auctane, and which of them are current lenders?");
console.log(r.state, r.answerState, r.answer);
for (const c of r.claims) console.log(c.claim_id, c.predicate, c.claim_type);
for (const url of r.evidenceUrls) console.log(url);
if (r.continuation) {
  const page2 = await research("", { continuation: r.continuation });
}

Public, runnable examples (anonymous research, citations, errors, pagination, a member key from the environment and a short tutorial), each with the output of a live run, are at https://github.com/Capital-W-Holdings/us-property-parcel-real-estate-debt/tree/main/examples/sdk.

The examples below live in the SDK source tree. Each example ran once, anonymously and with no paid API: 01 and 03 against production, 02 offline on the bundled receipt. Its real output is in .out.txt, and the UTC time of the run is in .ran_at. Build first with npm run build, then run it with node examples/<name>.

| example | investigation | |:-|:-| | 01_lenders_and_resolve.mjs | Which BDCs lend to Auctane, then resolve "Thoma Bravo" to DFX ids | | 02_read_receipt.mjs | Read a research receipt (dfx.receipt/2) with plain fetch; needs no build and runs offline on the bundled receipt (see ../RECEIPTS.md for the schema and which agent frameworks read receipts with no custom code) | | 03_ai_sdk_tool.mjs | The Vercel AI SDK tool executed the way generateText executes it (Hyland Software's BDC lenders), with no model call and no ai install needed |

The Hyland figures in example 03 were read against the ORTF 10-Q for 2026-06-30 and match (see the Python README, "Checked against the primary documents"). The Auctane answer in example 01 books a joint venture's loan to its parent BDC for Blackstone Private Credit Fund, a known DFX data issue listed in the Python README.

API

| call | transport | returns | |:-|:-|:-| | research(question, { continuation?, detail?, depth?, entity?, workflow?, as_of?, since? }) | MCP tools/call research on /mcp/core | ResearchResult | | verify(claim, structured?) | MCP tools/call verify | VerifyResult: SUPPORTED, PARTIALLY_SUPPORTED, CONTRADICTED or UNKNOWN | | resolve(name, { domain?, entity_type? }) | MCP tools/call resolve_name | ResolveResult: totalLowerBound means "at least N" | | delegate(question, { arguments?, contextId?, pollIntervalMs?, maxWaitMs? }) | A2A 0.3 message/send on /a2a, then tasks/get until final | DelegateResult with .result as a ResearchResult |

Use new Client({ apiKey, baseUrl, agentName, userAgentSuffix, channel }) to set options. Every result keeps the server's payload, unchanged, as .raw. receiptUrl is passed through when the server provides one and is null otherwise.

States, not exceptions

NO_MATCH, NOT_COVERED, UNKNOWN, RATE_LIMITED and INVALID_ARGUMENT arrive as result.state. The promise does not reject for them.

  • RATE_LIMITED comes with retryAfterSeconds.
  • NO_MATCH means the name resolved but nothing matches. Its near misses are in .raw.candidates.
  • UNKNOWN means there is no evidence either way. Absence of evidence is never a contradiction.
  • NOT_COVERED means DFX does not read that kind of record.
  • TEMPORARILY_UNAVAILABLE means the exchange is at capacity or not answering (HTTP 503). Nothing ran and nothing was charged; retryAfterSeconds is set when the exchange sent one.
  • API_KEY_INVALID, NOT_IN_PLAN and ENTITLEMENT_REQUIRED mean the key is not recognised or the plan does not cover the call; .raw carries how_to_fix or the plans URL.
  • HTTP_ERROR is any other non-2xx answer; .error carries the server's description.

Only transport failures throw TransportError, including a request with no answer within timeoutMs (90000 by default).

Free and member access

Anonymous (free, no key). All the read tools work, under DFX's published policy:

  • A list names its first 5 rows and only counts the rest.
  • A record names its subject and the first 3 related names per section.
  • Contact values and decision-maker names are never returned.

Anonymous use is limited per network address, not per agent: 20 calls a minute and 300 an hour, and the search and changes tools (research is one of them) share 60 an hour. Callers behind one office or cloud address share these budgets (budget.identity_basis network_prefix in the RATE_LIMITED envelope). agentName (X-Agent-Name) labels your agent; it does not give it a separate budget. A refused call is not charged and says when to retry.

Member. Pass apiKey or set DFX_API_KEY to a dfx_live_ account key. It goes out as Authorization: Bearer and unlocks withheld rows on your plan's verticals. A member is metered on the membership's own allowance in retrieval units a day and a month, not in calls: one research call uses 6 to 18 units depending on its depth. Allowances by plan: https://dfxintel.com/data-factory/plans

Withheld is not missing. When an answer says rows are withheld, they exist, and the answer counts them.

Coverage limits (stated by the service itself)

  • Private credit comes from BDCs' own SEC schedules of investments. Banks, CLOs and private funds that file no schedule are invisible, so principal is a lower bound.
  • Advisor moves use IAPD registration dates, not announcement dates.
  • Allocator commitments are the plan's own printed disclosures.

Each answer lists its own limits in coverageLimits and in .raw.do_not_conclude.

Framework adapters

| framework | status | |:-|:-| | Vercel AI SDK | shipped as dfx-intel/ai-sdk. Pass the ai module in, so this package still has no dependencies. Checked against ai 5.0.273, and on 2026-10-12 run inside ai 7.0.137 generateText (a scripted mock model called the tool, which answered from production). The offline tests use a stub. | | LangChain.js, Mastra | snippet only. Both consume DFX as a remote MCP server, so a wrapper would add nothing. |

import * as ai from "ai";
import { aiSdkTools } from "dfx-intel/ai-sdk";
const { text } = await ai.generateText({ model, tools: aiSdkTools(ai), prompt: "Who lends to Kaseya?" });

Mastra (@mastra/mcp):

import { MCPClient } from "@mastra/mcp";
const mcp = new MCPClient({ servers: { dfx: { url: new URL("https://exchange-production-9123.up.railway.app/mcp/core?c=sdk-ts") } } });
const tools = await mcp.getTools();

LangChain.js (@langchain/mcp-adapters):

import { MultiServerMCPClient } from "@langchain/mcp-adapters";
const client = new MultiServerMCPClient({ mcpServers: { dfx: { transport: "http", url: "https://exchange-production-9123.up.railway.app/mcp/core?c=sdk-ts" } } });
const tools = await client.getTools();

Endpoints and attribution

  • MCP: https://exchange-production-9123.up.railway.app/mcp/core?c=sdk-ts. A2A: /a2a on the same host.
  • Every request sends User-Agent: dfx-intel-ts/<version>. Browsers do not let a page set User-Agent, so they send their own.
  • Anonymous URLs carry ?c=sdk-ts. A member request (one with a key, or a URL ending in /member) is never tagged, because DFX's authorization server refuses an OAuth resource with a query string.
  • A2A tasks are held in server memory for up to 24 hours.

Tests

npm ci && npm test

The tests are offline. They run against recorded production responses in ../fixtures.

Licence

MIT. See LICENSE.