# Deal Sourcing Agent > Describe your investment thesis or name a company you like. Nimble agents search startup lists, funding news and hiring signals on the live web and return a ranked deal list with stage, investors, founders and a source for every value. Interactive demo: https://www.nimbleway.com/demos/deal-sourcing ## Find companies that match your thesis before anyone else Describe your thesis or name a company you like, and Nimble agents search startup lists, funding news and hiring signals on the live web for companies that fit. ## Examples to try - "Bootstrapped European B2B software, 20 to 200 employees" - Ramp, Rippling, Deel ## Set it up in your agent The Nimble skill for this use case: https://docs.nimbleway.com/skills/deal-sourcing/SKILL.md Prompt to paste into your agent: ```text Be my deal sourcing assistant using Nimble. Follow https://docs.nimbleway.com/skills/deal-sourcing/SKILL.md. ``` Get an API key: https://online.nimbleway.com/signup ## The most accurate & efficient web search for deal sourcing We asked Opus 5.5 to source companies for investment theses and to answer the questions an investor checks first: stage, ownership, last round, investors and founders. It ran once with each web search provider, and every answer was checked against a hand-verified ground truth. We tracked accuracy and the total cost of getting there, including search fees and tokens. Benchmark code and data: https://github.com/Nimbleway/deal-sourcing-benchmark ## What you get A ranked list of private companies that fit an investment thesis, researched on the live web. Each company carries its website, a one-liner, HQ, founded year, employees, stage and last round, investors, ownership, founders, a recent momentum signal, a fit score from 0 to 100, why it fits and why now. Stage, last round, investors, momentum and founders each link to a source. Stage and ownership show only when a source states them; otherwise they stay unknown, and a company that misses a hard criterion is left out instead of kept to fill the list. A short list is an honest answer. ## Modes | Mode | Input | What it does | |---|---|---| | A thesis | A sentence describing sector, geography, size and ownership | Finds companies that fit the thesis | | Lookalikes of a company | A company name or website, for example Ramp | Finds earlier-stage companies like that one | The demo then narrows by stage (Bootstrapped & founder-owned, Pre-seed–Seed, Series A–B, Growth, PE-backed), geography (Anywhere, United States, North America, Europe, UK & Ireland, Israel, Asia-Pacific) and list size (10 or 20 companies). It searches three angles in parallel: startup lists and directories, funding and deal news, and momentum (hiring, launches, awards). The list exports as CSV, one row per company with name and domain first, or as JSON. ## Inputs | Input | Required | Example | |---|---|---| | Thesis | Yes | Sector, geography, size and ownership in one sentence, for example "Bootstrapped European B2B software, 20 to 200 employees" | | Exclusions | No | For example "no PE-backed" or "not in our portfolio" | ## Run it yourself The Nimble skill for this use case runs the Nimble agent template `deal-sourcing`: https://docs.nimbleway.com/skills/deal-sourcing/SKILL.md - Create the agent once: `nimble_agents_create(template="deal-sourcing", agent_name="deal-sourcing")` - Run it: `nimble_agents_run(agent_name="deal-sourcing", input="Find companies in with . Exclude .", effort="medium")` - Example input: `Find bootstrapped European vertical SaaS companies for logistics or freight forwarding with 20-200 employees` - Effort: the skill runs the template at medium effort (about 1-3 minutes), even when the template's own default is different. Ask for low, medium, high or x-high to run another. State each criterion as a clear test, ownership most of all. For a big market map, split it by region or sub-sector, one run per slice in parallel, then merge and dedupe. Don't ask for a number of companies: narrow theses return few rows on purpose. Ownership is the hardest field, so spot-check the top rows before outreach. ## When to use it, and when not to Use it for "find acquisition targets", "deal sourcing", "map the market for", "founder-owned companies in", "off-market targets" or "companies that fit our thesis" requests. - For a sales prospect list, use Account Discovery: https://www.nimbleway.com/demos/account-discovery.md - To research one company in depth, use Due Diligence: https://www.nimbleway.com/demos/due-diligence.md ## Sample result From the demo's recorded run on the thesis "Vertical SaaS for construction in the US, Seed–Series A", thesis mode, stages Pre-seed–Seed and Series A–B, United States, 10 companies, recorded 2026-11-09. The recording does not state an effort level. The three angle agents found 11 matching companies; the top three: | Rank | Company | Stage & last round | Investors | HQ | Fit | |---|---|---|---|---|---| | 1 | Agave (useagave.com): AI agents that automate construction back-office financial work. | Series A, $15M, 2026-07 | Accel, Y Combinator | San Francisco, CA, United States | 92 | | 2 | Cascade (usecascade.ai): AI-powered business platform for AEC firms. | Seed, $3.5M, 2026-07 | Andreessen Horowitz Speedrun, Ada Ventures, Blitzscaling Ventures, Indico Capital Partners, Shucker VC, Snowball VC, G2C Ventures | New York, United States | 92 | | 3 | Buildcheck (buildcheck.ai): Computer-vision AI that reviews construction drawings for errors. | Series A, $12M, September 2026 | Telescope Partners, WND Ventures, Uncork Capital, Salt VC, Xfund | San Francisco, United States | 90 | Sources include https://useagave.com/blog/announcing-agaves-15m-series-a, https://techcrunch.com/2026/07/22/cascade-raises-3-5m-to-help-construction-firms-find-and-win-projects/ and https://www.fundraiseinsider.com/blog/buildcheck-raises-12m-series-a-for-ai-construction-drawing-reviews/ Each row also carries founders with a short background, why it fits, why now, and which angles found it. Fields no source stated, such as Buildcheck's founded year and employee count, come back empty and show as not published.