How to use Claude Code to build a signal-based prospecting list
The one-command signal-based workflow that saved tens of hours of manual research and scraping. Find recently funded companies that are about to grow.
AI agents are everywhere in GTM right now. LinkedIn is wall to wall with demos, YouTube is full of them, and the bigger companies are rebuilding their entire products around the trend. However, none of these come with the true numbers.
So when someone tells you to use Claude Code for lead generation, the fair question is whether it survives contact with a real client. This one did. It builds a signal-based prospecting list, meaning a target list assembled from live events instead of static filters, and the event here is a company raising money. The numbers above are from a single run.
I did not build it. Tanay Mishra did, for a performance marketing agency that was doing the job by hand. Scrape the news for companies that just raised, research each one, decide whether it is worth a pitch. It worked, and it burned tens of hours a week, which is the part nobody puts in the case study.
He is not affiliated with us and we did not pay for this. He built it, showed it in a video for the rest of the Clay Bootcamp members, and it is the cleanest version of a play I watch teams rebuild badly every month. Here it is step by step, including the parts where it falls over.
Watch the Claude Code workflow run
The walkthrough covers the whole build, from the filtered feed to the Slack notification. The guide below is the written version, with the steps broken out so you can follow along.
What is signal-based prospecting?
Signal-based prospecting means letting an event decide who enters your target list. Static list building starts from a firmographic profile, exports everyone who matches, and hands your team a list that was equally true last quarter. Signal-based prospecting starts from something that just happened, so a company enters the list the week it raises, hires, or loses an executive, and leaves it when the moment has passed.
That difference decides what the automation is worth. A workflow built on static filters produces the same list faster. A workflow built on events produces a list that did not exist yesterday, which is the only version worth putting on a schedule.
There is a second payoff and it shows up in the reply rate. Most outreach reads robotic because it is generalized. You pull a list of ICP-like accounts, you have nothing specific to say about any of them, so you send the message that could go to all of them and you turn into a spammer. The signal fixes that at the source. The event is the thing worth writing about, and it is different for every account on the list.
Claude Code sits in the middle of that. It is not the data and it is not the outreach tool. It is the operator that reads the event, researches the company, applies your judgement, and files the result where your team works. The rest of this guide is one worked example of exactly that.
Why recently funded companies are worth a list
A funding round is the highest-confidence moment in a company's year. His reasoning lines up with what we see in our own data every week, and with the research we ran across 474 funded startups on whether the funding signal is still worth anything in 2026:
- They have cash and a mandate to spend it
- They are about to hire
- They are under investor pressure to show returns quickly
- They are in the one window where they can take risks on new vendors
For an agency, every one of those is a reason to be in the conversation. Nobody struggles to identify this signal. The hard part is catching it while the window is open and doing the qualification fast enough that it still means something. That research put the median gap between a round and the company's first observable hire at 35 days, which is roughly how long you have. Miss by two weeks and you are pitching a company that already signed somebody else.
The Claude Code lead generation stack
Four tools plus a notification channel, each doing exactly one job.
The part worth taking from this is architectural. A workflow like this needs a real event detection layer at the front, and it is not something you can improvise with the tools next to it. Orchestration platforms poll the list you already gave them. Search APIs return what has been indexed, which is history by definition, and historical backfill is what they are genuinely good at. Ask either one to be the trigger and it will miss the freshness bar the whole play depends on.
Step 1: Narrow the funding signal before you automate it
Before you filter anything, it is worth knowing what sits behind the feed, because it decides what the rest of the workflow can do. We do not buy this data and we do not wait for a database to publish it. A round is caught from its first public trace, wherever that trace shows up:
- Social media, where founders and investors announce it before anyone writes it up
- Investor newsletters and portfolio pages
- News outlets, national and trade
- Law firms announcing the deals they closed
- Local newspapers, which is where a lot of smaller raises surface first
- Our own product logic layered on top of all of it
From there the company is resolved against the entity graph so the same raise does not arrive three times, and the profile is built around the event before it reaches you. Sourcing is the easy part. Verification and contextualization are what turn a raw mention into something you can act on, and all three arrive in one call.
Then open the funding feed and filter it down to what you actually sell to. He filtered by country, because his client sold into one market. The same filters work for round type, sector, city, and headcount.
Do this in the dashboard first, before you write a single line of the workflow. If the filtered feed does not already look like your target list, automating it just produces the wrong list faster.
Step 2: Point Claude Code at the API docs and the MCP
Grab an API key from your account, then hand Claude Code two things: the key, and the URL of the API documentation.
That second part is the trick that removes most of the work. Claude Code crawls the docs, learns the endpoints, the filters and the response shape, and writes its own integration. You never read the API reference yourself.
There is a shorter path, and it is where the real magic sits. The Signalbase MCP server exposes the same funding, hiring, and job change signals to Claude Code as native tools. No integration to write, no endpoints to learn, no HTTP calls in your workflow file. The agent queries signals the way it reads a file, and the same connection works in Cursor or any other MCP client you already run.
Step 3: Write the workflow file that does the listbuilding
Everything that makes this workflow yours lives in one markdown file. His is called workflow.md and runs with a single command. The file holds the instructions Claude follows on every run:
- Which signals to pull, and over what window (his is 14 days)
- Which filters to apply
- What your ICP is, in plain language
- What research questions to answer about each company
- How to classify fit
- Where to send qualified accounts, and what to do with the rest
Write the ICP section carefully. It is the part that decides everything downstream, and it is the part most people rush.
Step 4: Let Claude Code do the disqualifying
This is the step that separates the workflow from a glorified export.
For each funded company, Claude fires parallel Exa searches, reads what the company actually does, and scores it against the ICP in your workflow file. The run classified each company by the service that fits, routing B2B companies to outbound and D2C companies to paid ads.
Then it threw half of them away. 25 in, 11 out. That disqualification rate is the whole point, and it is the step most people skip when they build this themselves. A list of every company that raised money is not a target list. A list of the ones you can actually serve is.
Step 5: Push the qualified list to Clay and notify the team
Qualified accounts land in a Clay table with the fields the sales team needs: company name, raise date, amount raised, LinkedIn URL, the service category, and a written justification for why the account qualified.
Claude also drafts a pitch angle per account. Treat it as a research shortcut rather than as copy to send. He is explicit about this and he is right. The value is that a rep opens the account already knowing the angle, not that the message writes itself. If you want the copy side done properly, we wrote the signal-based copy playbook for exactly that.
A summary then goes to Slack so the team knows there is something to act on today.
Step 6: Put the Claude Code workflow on a schedule
The workflow runs on a cron job. Daily, weekly, or every 15 minutes if your market moves that fast.
Choose the interval based on how quickly your window closes. If you are pitching services that a funded company buys in month one, daily is the floor. If the buying cycle is slower, weekly is fine and cheaper.
Where this Claude Code workflow stops
Every workflow has a place where it falls over, and a guide that skips that part is a brochure. Here is where this one stops.
The qualification is only as good as the ICP you wrote in the workflow file. Vague input gives vague scoring, and that is the failure mode, not the tooling.
Treat it as a research shortcut that saves a rep an hour of reading. Sending it unedited is how you get the reply rate everyone complains about.
One country plus one sector plus one round type can return very few companies per week. That is correct behaviour, but plan the cadence around it.
Signalbase, Exa, and Clay are separate paid tools on top of Claude Code. Budget for the stack rather than for one subscription.
Adapting Claude Code prospecting to your business
The structure holds for any signal and any service. Swap the input, keep the shape.
A recruitment agency runs exactly this workflow with one change. The qualification question stops being "can we run their ads" and becomes "will they need to hire." Funded companies open roles within weeks of the round closing, which turns the same feed into a client-acquisition list rather than a marketing list. That version pairs funding with hiring signals and job changes on the recruiting data API, so the round tells you who is about to grow and the hiring signal tells you when they started.
The same applies to job changes for champion tracking, and to M&A for post-acquisition consolidation plays.
Start with the funding feed this workflow reads
A 7-day trial includes 500 credits, which is enough to run the full workflow end to end before you commit to anything.
Tested by 700+ GTM teams · No card charge
Claude Code lead generation FAQ
Go deeper
The research and infrastructure behind this workflow
The feed this workflow reads: rounds detected as they are announced, with the source attached.
474 funded startups, and the measured gap between the round and the first observable hire.
What to do with the pitch angle once the workflow hands it to you.