What comes with Edge Copilot, my Crawford research agent, the APIs in my stack and the cheapest ways I know to do research.
Included with the annual tier, $2,500/year.
I built each of these. You run them through Edge Copilot in Claude Code.
A Claude Code tool that builds production-ready Claygents for Clay, plus the course that teaches the method.
Asks you questions, checks Clay's native integrations first, tests against real data and scores with a 7-criterion rubric until the output is right.
Open AutoClaygentFinds confirmed customers of a technology, each with a URL you can click.
Fingerprints how a product shows up in DNS, certificates, headers and login pages, then scans for that fingerprint.
Find Every Customer of Any TechnologyTurns a screen recording of a list into clean rows.
Crops each row into its own image and reads them one at a time. Whole-frame reading got about 58% right; row by row got 99%.
597 Attendees From a 3-Minute Screen RecordingPulls every contact row out of directory PDFs.
Reads the text layer for free when it exists and crops rows only when it doesn't. Two directories gave 5,080 raw records at zero model cost.
The PDF Contact Extractor I BuiltBuilds a market list from public sources instead of a bought list.
Searches registries, licensing boards and trade associations for the vertical. One landscaping run surfaced 38 sources.
Stop Buying Lists. Build Them.One sorted timeline per account across CRM, calls, billing, product and support.
Stitches every system together and trusts what customers did first, what they said second and what the CRM says third.
The Dossier: Where AI GTM BeginsReads churned or lost accounts against clear hypotheses, using public evidence.
Checks each account in parallel and writes structured findings. On 1,200 churned accounts, 19% were never the right kind of operator.
Could We Have Seen the Churn Coming?Searches every plausible company for candidates instead of sampling a few.
Pulls every plausible person from every company in the filter, scores them and returns a ranked sheet with contact details.
Stop sampling candidates. Search them all.Runs one research question through ChatGPT, Claude and Gemini at the same time.
Checks every cited link, then leads with the claims only one engine found that survived the check.
Automated 3 Deep ResearchFinds a named contact for every company website in a market.
Sorts sites into free and paid buckets before spending anything. Two markets, 650 sites, no paid lookups.
A TAM of named contactsMatches company names to their domains at scale.
Free sources first, a cheap search next, a model only on ambiguous names, and every domain checked against what the site says.
Company name to domainFinds the best speaker for your event in any city.
Interviews you about the goal and audience, searches people data, stages, podcasts and your own network, then ranks finalists with a source behind every claim.
Speaker FinderThe six Claude Code skills I run on every engagement, free on GitHub.
A shared knowledge folder, a data auditor, an account dossier builder, won/lost analysis and an evidence packet builder, with a worked example to run first.
Get Started HereTurns five CRM exports into a Google Doc that grades every rep.
Follows each lead to its deal, compares reps inside one lead source and prices one coaching fix per rep in dollars.
Rep ReviewEvery U.S. business you sell to, ranked by recent customer complaints about the problems you fix, with the review to quote.
Reads every Google review from the last 12 months, marks the problems you confirm and hands back a ranked file with two messages to edit per business.
Open VerbatimCrawford is my research agent in Claude Code. It chases the cheapest dollar per right answer.
Its proving run got 34 of 39 hand-checked rows right.
Included with the annual tier.Introducing Crawford
The APIs and tools I use, each with the post where I wrote about it.
Stored people data. It answered in 0.3 seconds in my LinkedIn profile test.
Source: LinkedIn profile testContact data, including phone numbers, run through Clay.
Source: Phone provider testEnrichment tables and provider waterfalls. Credits I already own get spent first.
Source: Introducing CrawfordWhere I build every skill and run Crawford.
Source: Get Started HereWorks alongside Claude Code, each handing the other work.
Source: Dual Brain buildScores full profiles and reviews for a probability, cheap enough to read the whole market.
Source: Why System 1 models like JevLive LinkedIn profiles. My pick when a rep or a workflow is waiting on the answer.
Source: LinkedIn profile testFinds people who just changed jobs.
Source: LinkedIn profile testA public copy of the web. Checks whether a domain served real content, before any live request.
Source: Introducing CrawfordOpen place records with each business’s own website and phone.
Source: Introducing CrawfordFree lookups that return the official website Google has on file.
Source: Introducing CrawfordA second free source for official websites, checked against the first.
Source: Introducing CrawfordWays I drive research costs down. Do the free steps first and pay only where you have to.
Instead of
Paying a scraper service to pull restaurant websites.
Do this
Have Claude Code download every restaurant website to your computer, for free.
Instead of
Having an AI model read every website.
Do this
Run a Python script that downloads the sitemap XML. It lists every page Google knows about, so you find the events page for free.
Put it together
Download the events pages for free, then check them for keywords with Python or run Jev on each one. You can size the event space of every restaurant in the country for about $10.
This is the Jevons paradox at work. Better, cheaper tools mean far more is possible. See what that makes possible
