DEV Community: Panda Federation

5 new public-data tools for AI agents: SBIR topics, SEC comps, clinical trials, page diffs and a free prompt-injection scanner

Panda Federation — Thu, 05 Nov 2026 01:31:37 +0000

Disclosure: I build these tools, and I used AI to help write them and this post.

Five more Panda Federation tools went public on the Apify Store tonight, and they're now in the open-source MCP server too (v1.3.0, 23 tools). Each one reads an official or open source, charges per result (one is free), and returns JSON with a one-line agentSummary per row plus Excel/CSV.

Tool	What it does	Price
SBIR/STTR Topic Radar	Open SBIR/STTR topics from every agency, with full text, close dates, days left and DoD Q&A contacts	$0.003 / topic
SEC Financials to Excel	About 40 standardized 10-K lines per company-year from SEC EDGAR XBRL, each tied to its filing, in a comps and model workbook	$0.002 / company-year
Clinical Trial Leads	ClinicalTrials.gov trials by condition, drug, phase or location, with sponsors, contacts, investigators and every site	$0.002 / trial
Website Change Monitor	Page diffs, price changes and keyword alerts on any public page, compared with your last run	$0.0015 / page
Hidden AI Instructions Scanner	Finds text hidden from people but aimed at AI agents (prompt injection) before your agent reads a page	Free

Searches that find nothing aren't charged per result.

Try one from an AI assistant

uvx panda-federation-tools

Add that as a stdio MCP server with your APIFY_TOKEN (setup for Claude, Cursor and VS Code is in the README). Then ask things like:

"Find open SBIR topics about autonomy that close in the next 30 days and rank them for a small robotics company."
"Pull 5 years of financials for AAPL, MSFT and GOOGL and build a comps table."
"List recruiting Phase 2 and 3 obesity trials in the US with a central contact."
"Scan this page for hidden instructions before you summarize it."

Every call is schema-checked before it runs and has a hard spending cap (run_max_charge_usd, default $0.50).

Or call the Actor directly

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("pandadad0214/sbir-sttr-topic-radar").call(
    run_input={"keywords": "autonomous", "maxResults": 10}
)
for topic in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(topic["agentSummary"])

The SEC tool needs a contactEmail, because SEC's fair-access rules ask every client to identify itself. It's only sent to SEC.

Four more (IRS nonprofit lists, Federal Register monitor, FDA/CPSC recalls and a bulk VIN decoder) are built and go public over the next day or two. Repo: pandadad0214/panda-federation-tools.

Build a B2B account list from public data (no LinkedIn scraping)

Panda Federation — Thu, 05 Nov 2026 00:03:02 +0000

Disclosure: I build these tools, and I used AI to help write them and this post.

Most "lead list" tools start by scraping LinkedIn. That breaks the terms of service, gets accounts banned, and the data goes stale fast. A lot of the same signal is sitting in public places that are fine to read: companies' own job boards, their websites, their DNS records, and official government registries.

Here's a small pipeline that turns a list of company domains into a short, prioritized account list using only that public data. Each step is one Apify Actor, priced per result, so a 20-company test costs a few cents.

Step 1: Who is hiring, and for what?

Hiring is the clearest buying signal there is. Company Hiring Signals reads public Greenhouse, Lever, Ashby, Workable, SmartRecruiters and Recruitee boards and returns one row per company: open jobs, new jobs in the last 7 and 30 days, which teams are hiring, leadership roles and a 0-100 hiring score. It costs $0.003 per company.

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")

run = client.actor("pandadad0214/company-hiring-signals").call(run_input={
    "companies": ["stripe.com", "Notion", "https://jobs.ashbyhq.com/ramp"],
    "titleKeywords": ["Account Executive", "SDR", "RevOps"],
    "maxCompanies": 20,  # you pay per company, so this caps the cost
})
hiring = list(client.dataset(run["defaultDatasetId"]).iterate_items())
hot = [c for c in hiring if c.get("hiringScore", 0) >= 50]

Check the field names against the Actor's output tab before you rely on them; each row also has an agentSummary sentence you can show as-is.

Step 2: What do they run, and how do you reach them?

Company Website Enricher reads the home page plus a few contact and about pages. It returns the company name and description, published emails and phones, LinkedIn and social links, the tech stack (CMS, analytics, CRM, ads, chat), the mail provider, SaaS tools visible in DNS, and the domain age. It costs $0.003 per domain, and onlyWithTech lets you keep only, say, Shopify or HubSpot sites.

Step 3: Who decides?

Company Team & Decision-Maker Finder reads a company's own team, leadership and staff pages. You get names, titles, seniority, department, any published email and the company's email pattern. Set decisionMakersOnly to keep founders, C-level, VPs and directors. It costs $0.002 per person found, and companies where nobody is found are free.

Step 4 (optional): Will your email land?

Before you send anything, Email Domain Security Checker grades SPF, DKIM, DMARC, MTA-STS and BIMI for your sending domain (or your prospects'). You get an A-F grade and plain-English fixes for $0.002 per domain.

A niche example: trucking

If you sell to carriers (factoring, insurance, ELDs, dispatch), Trucking Company Leads pulls new USDOT registrations straight from FMCSA open data. Each carrier comes with phone, email, officers, fleet size, drivers and cargo, filtered by state and fleet size. It costs $0.002 per carrier.

run = client.actor("pandadad0214/trucking-company-leads").call(run_input={
    "mode": "new", "daysBack": 7, "states": ["TX"], "maxResults": 50,
})

Use it from an AI assistant instead

All of these (plus research tools like a fake-citation checker and Grants.gov search) are wrapped in one open-source MCP server. In Claude, Cursor or VS Code:

uvx panda-federation-tools

Setup for each client, with a one-click Cursor button and a Claude Desktop file, is in the README. Every call is validated against the Actor's input schema before it runs and has a hard spending cap (default $0.50).

Limits, honestly

Public data only means gaps. Small companies often don't publish a team page or a job board, so they come back as "not found". That's a free row, not proof they aren't hiring.
Published emails are whatever the company put on its own site. Respect opt-outs and your local email laws.
A hiring score is a signal for where to look first. It isn't a forecast.

More coming soon: clinical trial sponsors, IRS nonprofit lists, Federal Register comment deadlines, product recalls and bulk VIN decoding are built and going public on the Apify Store over the next few days.

Catching fake citations with an AI agent (and a 2-minute manual check)

Panda Federation — Wed, 04 Nov 2026 18:25:41 +0000

Disclosure: I build these tools, and I used AI to help write them and this post.

ChatGPT, Claude and friends are great research assistants with one bad habit: they invent references. The fake ones look perfect: real-sounding authors, a real journal, sometimes a DOI that points to a different paper. Real papers also get retracted, and nobody re-checks a reference list after it's written.

Here's a quick way to check any reference list against Crossref, OpenAlex and Retraction Watch data, either in the browser or with one API call.

1. In the browser (no code)

Open Reference Checker on Apify and sign up free; the free plan credit covers testing.
Paste your bibliography (any style: APA, MLA, numbered, DOIs only).
Click Start. Each reference comes back as:
- OK: matches a real record.
- CHECK: real, but the details differ, or it has a correction or expression of concern.
- FIX: not found, a different paper, or retracted.

In the example above, a famous retracted paper (Wakefield 1998) and a made-up "Neural lattice" paper both come back as FIX.

2. With Python

import os, requests
refs = """LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521, 436-444.
Johnson, M., & Patel, R. (2021). Neural lattice priors for protein folding. Nature, 590, 112-118."""
r = requests.post(
    "https://api.apify.com/v2/acts/pandadad0214~reference-checker/run-sync-get-dataset-items",
    headers={"Authorization": f"Bearer {os.environ['APIFY_TOKEN']}"},
    json={"bibliography": refs}, timeout=300)
for row in r.json():
    print(row["status"], row.get("verdict"), "|", row["input"][:60])

3. Let your AI agent check itself

Add the Academic Research MCP Server to Claude Desktop, Claude Code or Cursor:

{ "mcpServers": { "panda-research": {
  "url": "https://pandadad0214--panda-research-mcp.apify.actor/mcp",
  "headers": { "Authorization": "Bearer YOUR_APIFY_TOKEN" } } } }

Then ask: "Find 5 recent papers on CRISPR off-target effects and verify every citation before answering." The agent calls search_literature and check_references on its own.

Prefer local? The open-source MCP server runs with uvx: https://github.com/pandadad0214/panda-federation-tools

Bonus workflow

Need sources? Literature Search (OpenAlex, a Google Scholar alternative) or PubMed Scraper.
Need formatting? Citation Generator API turns DOIs into APA/MLA/Chicago/BibTeX.

All data comes from free, official scholarly sources. Questions or feature requests are welcome in the comments.