Signal-based prospecting in Claude Code means asking an agent to find companies that just did something relevant (raised a round, opened a role, adopted a tool), confirm the signal, find the right person and write outreach that references it. With the Databar MCP connected, each step maps to a specific enrichment you can name, price and rerun every week.
The strategy behind signals is covered elsewhere on this blog, in signal-based GTM and the 10 best buying signals. This post is the build: which Databar enrichments to call for funding, hiring and tech-stack signals, three plays with prompts you can paste, and how to rerun them on a schedule.
What makes a signal worth acting on?
A signal earns outreach when it passes three checks. It is recent enough that the situation it describes is still live. It is relevant to the problem your product solves, not just "news". And it points to a reachable person who owns that problem. A funding round at a company outside your ICP fails the second check. A tech-stack match with no identifiable owner fails the third. Put these checks in your instructions so Claude drops rows instead of writing weak emails about them.
Which Databar enrichments cover funding, hiring and tech-stack signals?
These come from the Databar catalog as of September 2026. Prices are credits per call, and only calls that return data are charged. The agent can confirm current prices with search_enrichments and get_enrichment_details before running anything.
Job | Enrichment (ID) | Source | Input | Credits |
|---|---|---|---|---|
Find companies by funding, size, industry or technology | Search companies with filters (677) | Filters such as min_funding_usd, employee range, company_technology_slug_or | 7.5 per page of 5 companies | |
Confirm funding events and dates | Get financing events of a company (679) | Domain | 7 | |
Open roles at a company | Job postings waterfall (job_postings_getter) | Leadmagic, PredictLeads | Company website | 1.5 or 7 |
Tech stack of a known company | Get a company's tech stack (977) | TheirStack | Domain | 6 |
Person by title at a company | Find people by company and job title (887) | Leadmagic | Domain and title | 6 |
Verified work email | Email by name waterfall (email_getter) | Up to 9 providers, plus a verifier | First name, last name, company | From 3, plus 1 to verify |
Other sources sit in the same catalog (Owler and Crustdata for funding, Limadata and Pubrio for jobs, BuiltWith and Pubrio for tech stack), so if one provider's data looks thin for your market, the agent can swap it without new keys.
How do you set up Claude Code for signal work?
Connect the Databar MCP with your API key (workspace, then Integrations):
Then give Claude your signal rules in the project's CLAUDE.md, so every play applies the same filters:
The windows and titles above are examples. Set them from your own closed-won history; the funding-signal guide discusses how timing affects response.
Play 1: How do you prospect companies that just raised?
The workflow the agent runs:
Call Search companies with filters (677) with your size, location and funding filters. The response includes lastfundingrounddate and lastfundingroundamountreadable, so Claude can drop anything outside your window.
For the survivors, confirm the event with Get financing events (679), which returns the financing type, date and a source link.
Find the owner with Find people by company and job title (887).
Get a verified email with the email_getter waterfall and a verifier.
Write rows to a Databar table with create_table and create_rows, so you can review them in the browser and export them.
Prompt:
A good opener names the round and connects it to a concrete consequence ("hiring your first AEs after the Series A usually means..."). A bad one congratulates them and pitches.
Play 2: How do you turn hiring signals into outreach?
Hiring works best on an account list you already trust, because a per-company job lookup is cheap (from 1.5 credits) and you already know the accounts fit. Start from your target accounts CSV or CRM export:
Run the job_postings_getter waterfall on each company website. It returns job titles, links, descriptions and locations.
Have Claude classify the roles against your signal rules. Three open SDR roles and a RevOps role is a different story from one backfill.
Find the hiring manager's likely title (for SDR roles, the sales leader) with enrichment 887.
Write outreach that references the specific role and what that hire will need on day one.
For more on reading job data, see how to use hiring signals and monitoring competitor job postings.
Play 3: How do you target companies by their tech stack?
There are two ways in, and they measure different things:
Discovery: Search companies with filters (677) accepts company_technology_slug_or, which matches technologies a company mentions in its job postings. Good for finding companies that work with a tool daily, such as a CRM or data warehouse.
Check a known list: Get a company's tech stack (977) returns the stack for a domain you already have. BuiltWith's lookup (170, 12 credits) detects technologies on the website itself, which suits marketing and web tools.
Job-posting mentions show what a team uses or plans to use; website detection shows what is installed on the site. Neither proves who owns the tool, so the contact step still matters. The agent can ask get_param_choices for enrichment 677 to find the exact technology slug before searching. Technographic targeting goes deeper on picking the right technologies.
How do you rerun signal plays every week?
Signals go stale, so the value is in running the same play on a cadence. Three options, from least to most setup:
Save the play as a skill. Put the prompt in .claude/skills/funding-signals/SKILL.md and run /funding-signals each Monday.
Run Claude Code headless. claude -p runs a prompt non-interactively, so a cron job or CI schedule can trigger the skill. Pre-approve only the Databar tools the play needs with --allowedTools, for example mcp__databar__run_bulk_enrichment and mcp__databar__run_bulk_waterfall.
Move the logic into Databar. Keep target accounts in a Databar table and set the job-postings enrichment to run on a schedule (scheduling options depend on your plan), with run conditions so only matching rows are processed. Multi-step logic can live in a Databar Flow, which your agent can also call as a tool.
Whichever you pick, dedupe against last week's output before enriching contacts, so you don't pay twice for the same account.
MCP, CLI or SDK for signal sweeps?
Use the MCP while you design a play: the agent can inspect enrichments, test filters and show you rows as it goes. Once the play is stable and runs over hundreds of accounts, move the bulk steps to the Databar CLI or Python SDK, where a CSV goes in and a CSV comes out without filling Claude's context. The MCP vs SDK vs API guide covers the trade-offs.
Surface | Pricing model | Best for | Data coverage focus | API/MCP access |
|---|---|---|---|---|
Databar MCP in Claude Code | Credits per call; no charge when nothing is returned | Designing and testing plays, small weekly batches | Full catalog: funding, jobs, tech stack, people, emails | Hosted MCP |
Databar CLI or Python SDK | Same credits | Scheduled sweeps over large account lists | Same catalog | CLI, SDK and REST API (plan-dependent) |
Databar tables and Flows | Same credits | Recurring runs your team can inspect without a terminal | Same catalog | Runs in the app, by API or as an agent tool |
Where does signal-based prospecting break?
Stale signals. A round from last year is not a trigger. Always keep the date column and filter on it.
False positives. Recruiting agencies post many jobs; the company search response includes an isrecruitingagency flag, so filter on it.
Weak links between signal and pitch. If you can't say in one sentence why this signal creates the need you solve, skip the account.
Over-personalization. Citing a job post is useful. Listing everything the agent found about a person is not.
For a wider view of agent-driven signal pipelines, see signal discovery with agents.
FAQ
Do I need separate subscriptions to TheirStack, PredictLeads or Leadmagic?
No. Through Databar they run on your Databar credits with one API key. If you already have your own key for a supported provider, you can connect it instead.
How many credits does one funding play use?
It depends on how many companies survive each filter. Searching costs 7.5 credits per page of five companies, confirming a round 7, finding a contact 6, and a verified email from 4. Filter hard before the contact and email steps, because that is where cost per account adds up.
Can Claude Code monitor signals continuously?
Not on its own. Claude Code runs when you or a scheduler start it. For recurring checks, trigger a headless claude -p run on a schedule or set scheduled enrichments on a Databar table.
Does this work in Cursor or Codex?
Yes. The Databar MCP works with any MCP client, including Cursor, Codex and Gemini CLI. The prompts above work the same way; only the project instruction file differs.
Run your first signal play
Pick one play, run it on 20 companies and read every row before scaling. Start free with the 14-day trial and 100 credits, or book a founder demo to map the signals that fit your product. The Databar MCP page has the connection details for other clients.
Recent articles
See all
