Startup thesis search

Skill jongall45/frontrun-skills/skills/startup-thesis-search

Agent skills for sourcing startup deals with Frontrun — pre-round signal from 1,000+ investor follow graphs on X. For Claude Code, Cursor, and any MCP-capable agent.

Install

npx -y skills add jongall45/frontrun-skills --skill startup-thesis-search

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

22 days oldThe repository was created 22 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Semantic startup search by investment thesis with the Frontrun API: describe a thesis in natural language ("agent infrastructure for regulated industries") and get matching early-stage companies ranked by meaning, each with first-flagged provenance and investor-convergence counts. Use when asked to find startups matching an investment thesis, run a semantic or natural-language company search, source companies in a specific theme or category, or map a market of early-stage startups.

SKILL.md

4.3 KB, as published. Nobody here has run it

Thesis search: natural language in, early-stage companies out

GET /v1/search/thesis embeds your thesis and matches it against embedded company descriptions by meaning, not keywords. "Platforms where people trade on what they know" finds prediction markets that never say the words "prediction market".

Auth and pricing (the honest version)

Every /v1 request needs an API key in the X-API-Key header. Keys are generated at https://www.frontrun.vc under Settings > API Keys. API + MCP access is included with the Frontrun Pro plan ($99/mo, 10,000 monthly credits; pay upfront, API access unlocks immediately, no trial on Pro). The Frontrun app itself has a 7-day free trial, but the API and MCP surface sit behind the Pro paywall. Top-up credit packs never expire.

Every charged response returns cost_credits and balance_credits. A 402 means insufficient credits; a 429 means a rate limit or the daily credit cap. Several expensive endpoints never charge for empty answers (convergence, thesis search, classify, founders, funding, similar VCs return cost_credits: 0 when nothing is found).

No key yet? Two free no-auth surfaces: GET https://frontrun.vc/v1/teaser/trending (JSON, top 5 trending companies, counts only) and the public page https://frontrun.vc/trending/. Both are real live data.

Machine-readable references: https://www.frontrun.vc/llms.txt and https://www.frontrun.vc/openapi.json

The endpoint

GET /v1/search/thesis

Semantic search over your database (companies surfaced by the accounts you track, not the global catalog). Describe an investment thesis in plain text and get companies whose descriptions match the meaning, not the exact words, ranked by cosine similarity. q must be at least 10 characters or the call fails with 400. Costs 40 credits; zero matches cost 0 credits. Each result includes first_flagged_at (when Frontrun first flagged the company) and convergence_count: cite these as provenance for how early the signal is.

Parameters:

q (query, required) Investment thesis in plain text, at least 10 characters (e.g. "information markets - platforms where people trade on what they know").
limit (query, default 25) Maximum results to return (default 25, max 50).

curl -s "https://frontrun.vc/v1/search/thesis?q=information%20markets%20where%20people%20trade%20on%20what%20they%20know&limit=10" -H "X-API-Key: $FRONTRUN_API_KEY"

What makes the results useful

Each match carries the receipt layer:

similarity: cosine similarity to your thesis, 0 to 1 (results come back ranked by it).
first_flagged_at: when Frontrun first flagged the company. Cite it as provenance.
convergence_count: how many tracked investors converged on it.
followers_count, sector, website, description.

Present matches with receipts, e.g. "@prlnet flagged 2026-03-17, now followed by 153 tracked investors".

Keyword fallback and filter discovery

Thesis search is scoped to your database (companies surfaced by accounts you track). For exact-string matching over the same graph use GET /v1/search (4 credits, requires at least one of sector, keyword/q, or entity_type, otherwise 400). Discover valid sector strings with GET /v1/sectors (4 credits).

curl -s "https://frontrun.vc/v1/search?keyword=prediction%20markets&limit=10" -H "X-API-Key: $FRONTRUN_API_KEY"
curl -s "https://frontrun.vc/v1/sectors" -H "X-API-Key: $FRONTRUN_API_KEY"

Recipes

search my thesis: "stablecoin infrastructure for emerging markets" - top 5 matches
with founders and one-line pitches.

map the agent-infrastructure landscape: thesis search 3 phrasings, dedupe, rank by
convergence_count, then pull funding for the top 10.

Pair with startup-fundraising-data to enrich matches (founders, funding) and track-investors-on-x to monitor the theme going forward.