Orange Collective
Financial Datasets

Financial Datasets

Connect your agent to the stock market

financialdatasets.aiYYC ProfileGitHub@findatasetsLinkedIn
Financial Datasets · YC S26 launch video

3,000+

Paying agents, 10 weeks after launch

All self-serve, no sales team

#1

Professional investors are the largest segment

Their agents call the API continuously during the trading day

1–2s

Filing latency after SEC publication

Structured 10-K / 10-Q via XBRL

27,000+

US tickers, 30+ years

98% self-sourced; global names since Aug

Thesis

The market data industry took in a record $49.2B in 2025, almost all of it sold as per-seat subscriptions; a Bloomberg terminal costs about $32K a year.[1][2] Analysts now do research inside Claude Code, Codex and agent harnesses their funds build, and those tools need structured data with source citations and pay-as-you-go pricing, not a terminal seat.[3][4][5][6] Financial Datasets parses SEC filings itself and serves structured data for 27,000+ US tickers over 30+ years, available within seconds of publication, priced from $0.01 per request.[7][8][9] Ten weeks after launch it has 3,000+ paying agents, all self-serve, and professional investors are the largest customer segment because their agents call the API continuously.[8] Agents are becoming a separate customer for market data, paying per request rather than per seat, and Financial Datasets is the data source many of them are already built on. Once a fund's agents are built on its API, switching is costly.

  1. 01

    Professional investors are the largest segment. Their agents call the API continuously during the trading day, so a fund's agents generate far more requests than an individual developer's.[8]

  2. 02

    The buyer is the fund's AI team, not procurement. Hedge funds are forming small teams of two or three data scientists and an analyst to build agents. Those teams sign up self-serve, and their usage is the path to a fund-wide contract.[8]

  3. 03

    Incumbents only sell to existing subscribers. FactSet, S&P, LSEG and Bloomberg give agents access only through subscriptions the firm already holds.[10][11]

  4. 04

    The data is self-sourced. 98% is crawled and indexed in-house, the XBRL (eXtensible Business Reporting Language) taxonomy is mapped by hand, and gross margin is above 70%.[8][7]

Problem

SEC filings are free, but normalizing them is hard.

Bloomberg, FactSet and S&P built their products for analysts at terminals, so they sell screens, exports and per-seat licenses.[6] The SEC publishes filings for free,[12] but companies report against about 17,000 different taxonomy variants,[8] and no vendor sells the normalized data with pay-as-you-go pricing. A developer building a finance agent has three options: normalize filings themselves, buy a vendor's statements on a monthly subscription with a request quota,[13][14] or use free feeds.[15] Data quality matters. In one vendor's benchmark, a frontier model answered financial questions correctly 91% of the time with a structured database and 20% of the time with web search.[16]

Why Now

Agents became a buyer of market data in the last eighteen months.

Models can parse filings, and agents now run long enough to use the data.

Frontier models can now extract structured data from an 8-K or an earnings call transcript in under a minute at acceptable cost.[8] The incumbents have started serving agents: FactSet, S&P, LSEG, Morningstar and Bloomberg all launched MCP (Model Context Protocol) servers in the past nine months,[10][17][18][19] and FactSet had 450+ clients on its server by July, with API volume up 13x in one quarter.[20]

Agents now run long enough to be their own data consumers. Anthropic, OpenAI and Google each launched finance agent products this year,[3][21][22] about 40% of Anthropic's top 50 customers are financial institutions,[23] and JPMorgan runs agents for hours on tasks that used to take minutes.[24] But every incumbent's agent access requires an existing subscription,[10][11][19] so a fund without one cannot buy filings-grade data pay-as-you-go.

Product & Traction

One structured index of filings, prices and events, sold pay-as-you-go.

Search, Answer and Monitor on one index, priced from $0.01 per request.

Financial Datasets crawls filings, prices and events into one structured index that updates every few seconds.[8] Structured 10-K and 10-Q data is available 1 to 2 seconds after the SEC posts it; earnings call transcripts and 8-Ks are extracted with frontier models in 20 to 35 seconds.[8] The API covers fundamentals, KPIs and guidance, filings, ownership, prices and news, delivered over REST, an MCP server and webhooks.[7][25] Three agent-facing functions sit on top: Search (natural-language queries returning structured results), Answer (direct answers to questions) and Monitor (event alerts pushed to long-running agents).[8] Pricing is pay-as-you-go from $0.01 to $0.10 per request, with redistribution included and no per-seat fees.[9][6]

Traction (launch post, founder call)[6][8]

Traction
  • 3,000+ paying agents on the platform ten weeks after launch, all self-serve; professional investors are the largest customer segment[8]
  • 1,000+ customers and millions of requests a day, with no sales team[6][8]
  • Gross margin above 70%[8]
  • The founder's open-source agents ai-hedge-fund (63.2K GitHub stars) and dexter (27.6K) both require a Financial Datasets API key[26][27]
  • Exa, Parallel and Firecrawl each asked to license the index and were turned down.[8] Exa now lists Financial Datasets as a pay-per-call data provider in Exa Connect[28]

There is no independent press coverage yet. The one third-party comparison rates FMP ahead on breadth and price and recommends Financial Datasets for filings data.[29]

Market

Sized from the seat budget and the number of agents that will consume data.

The industry spent $49.2B on market data in 2025,[1] nearly all of it per seat; Bloomberg alone has about 325,000 terminals at $31,980 a year.[30][2] There is no published market size for data consumed by agents. IDC projects more than 1.3B deployed agents by 2028.[31] Spend per agent grows as agents run longer tasks, and a professional investor's agents spend far more than an individual developer's.[8] The nearest comparable is the developer data-API market, priced at $29 to $2,000 a month, where every vendor still sells monthly subscriptions with request quotas rather than pay-as-you-go pricing.[13][14][32]

Market map & competitive landscape

No other vendor sells primary-source financial data pay-as-you-go.

Financial Datasets sits between two groups. Licensed incumbents and primary-source data extractors hold normalized financial data but sell it per seat or by custom quote. Exa, Parallel and Firecrawl were built for agents and already offer pay-as-you-go pricing, structured output, citations and webhook monitoring, but none has a financial data schema. Their results cite web pages, not XBRL facts, and their financial data comes from EDGAR PDFs or from partners they pay.

Exa

Resells Financial Datasets today

Sells web search, answers and monitoring pay-as-you-go. Raised a $250M Series C led by a16z at a $2.2B valuation in May and counts private-equity and consulting firms among 5,000+ customers.[33][34][35] Its citations are web URLs, not filings. Since June it has resold Financial Datasets as a pay-per-call provider in Exa Connect, so it is currently a distribution channel, not a competitor.[28]

Parallel

Closest on structured output

The closest of the three: its Task API returns schema-conformant JSON with per-field citations, its Monitor API pushes events by webhook, and pricing is pay-as-you-go. Raised a $100M Series B led by Sequoia at a $2B valuation in April; customers include Rogo and Harvey.[36][37][38][39] Its sources are web pages and EDGAR PDFs, and its financial data comes from partners it pays through its Index program, including Fiscal AI and PitchBook.[40][41] That makes Parallel a likely licensee of a financial schema, and the best-placed company to build one if it chooses.[42]

Firecrawl

Returns web pages

A YC web-scraping API priced per page. Raised a $14.5M Series A led by Nexus in 2025 and reports 1.25M+ developers and eight-figure ARR in its first year.[43][44][45] It returns web pages rather than structured financial data, describes itself as complementary to fundamentals APIs, and reviews Financial Datasets as "one of the most thoughtfully built options".[46]

Everything else

Requires an existing subscription

Bloomberg, FactSet, S&P/Kensho, LSEG and Morningstar give agents access only through an existing subscription.[10][11][19] Daloopa ($47M Series C in May) sells human-verified filings data by custom quote and is listed on the Anthropic, OpenAI and Google connector directories; sec-api.io and edgar.tools sell raw XBRL cheaply with no KPI extraction.[16][47][48] Developer APIs such as Massive, FMP, Intrinio and Databento are in the same price range but resell vendor data on monthly quotas.[13][14][32][49]

Competitive landscape, table view

CategoryExamplesPosition
Horizontal retrieval: ExaSearch, Answer, Research, Monitors, Exa ConnectSame pricing model and interface, far more capital, no financial schema.[33][35] Currently a channel through Connect;[28] becomes a threat if it builds or buys its own financial data.
Horizontal retrieval: ParallelSearch, Task, Monitor, IndexClosest on structured output and event push.[36][38] Pays partners for financial data;[41] a likely licensee today, and the best-placed builder if it chooses.
Horizontal retrieval: FirecrawlScrape, Crawl, Search, Agent, MonitorCheapest per unit and the largest developer base, but returns web pages.[43][46] Overlap is limited to agents scraping EDGAR directly.
Licensed incumbentsBloomberg, FactSet, S&P/Kensho, LSEG, Morningstar/PitchBookThe deepest data, listed on every platform's connector directory,[50][51][22] but only for existing subscribers.[10][11] A direct threat to pro-investor revenue if those desks reach the same data through subscriptions they already hold.
Primary-source extractorsDaloopa, sec-api.io, edgar.tools, Fira MCP, Fiscal.aiSame data model. Daloopa leads on verification, distribution and capital and sells by quote;[16][47] the others stop at raw filings.[48]
Developer market-data APIsMassive, FMP, Alpha Vantage, EODHD, Intrinio, DatabentoSame price range and self-serve, but none parses filings itself or prices pay-as-you-go.[13][14][32] Substitution risk on prices and statements only.
Agent platforms as channelAnthropic, OpenAI, Google finance productsThe distribution chokepoint. All three connector directories list FactSet, S&P, LSEG, PitchBook and Daloopa; none lists Financial Datasets.[50][51][22]

Assessment

To become the default data source for agents, a vendor needs primary-source data in one schema, an agent-native interface, and distribution where agents are built. Exa, Parallel and Firecrawl have the interface and more capital, and two of them already sell the monitoring product Financial Datasets calls its third function.[35][38] None has the financial schema, and the two that handle financial data pay other providers for it.[28][41] Financial Datasets has the schema and the open-source developer funnel.[8][26][27] Two things have to hold: Exa, Parallel and Firecrawl keep buying the data rather than building their own financial schema, and professional investors keep paying self-serve once FactSet and Daloopa are available inside the same agents through their existing subscriptions.

Founders

Virat Singh

Virat Singh

Founder

Buy-side M&A, then fourteen years as a software engineer: search lead at Faire (YC), market-data pipelines at Acorns, search and payments at Airbnb. First-time founder.

Risks

Risk

Exa and Parallel already sell pay-as-you-go search with structured output and monitoring, and raised $250M and $100M this year.[33][37]

Mitigation

Both pay other providers for financial data rather than build their own;[28][41] the risk stays contained as long as mapping 17,000 taxonomies is a worse use of their engineers than paying Financial Datasets per request.

Risk

Raw filings are free from the SEC,[12] so the paid layer is normalization, KPI extraction and speed, which Daloopa already sells with human verification.[16]

Mitigation

A gross margin above 70% shows buyers pay well above the free floor;[8] the position holds as long as Daloopa keeps selling by seat and quote.

Risk

Coverage is US equities only; the global and cross-asset data the system-of-record thesis needs is still on the roadmap.[8]

Mitigation

The schema extends to each new market at the cost of mapping it, and the second-half international rollout needs to ship on schedule.

What we're watching

  • Usage revenue growing more than 15% a week through Q4 as legacy subscriptions move to pay-as-you-go.[8]
  • The four New York engineers hired by year end and the open-source repositories maintained again.[52]
  • When the company adds a top-down sales motion for hedge funds. The plan is to stay self-serve until after the Series A, since fund sales cycles run 45 to 60 days with heavy compliance review.[8]
  • A listing on an Anthropic, OpenAI or Google finance connector directory.[50][51][22]
  • International coverage live at the same latency, Exa Connect volume, and whether Parallel's Index adds Financial Datasets.[8][28][41]
  • Expansion beyond US equities into adjacent markets, starting with crypto and prediction markets, which the founder has on the roadmap after international equities.[8]

References

  1. [1]Real-Time Analytics Drive Record Growth Across the Financial Market Data Landscape (Burton-Taylor via Mondo Visione, March 2026)
  2. [2]Bloomberg Terminals: How Much More You'll Pay Next Year (NeuGroup)
  3. [3]Agents for financial services (Anthropic, May 5, 2026)
  4. [4]NBIM: extending Claude Code to business analysts and quantitative researchers (Anthropic customer story, 2026)
  5. [5]OpenAI launches Codex plugins for finance pros (Finextra, June 2, 2026)
  6. [6]Launch YC: Financial Datasets: Stock market data for AI agents (August 2026)
  7. [7]Introduction, Financial Datasets docs
  8. [8]Financial Datasets founder call with Orange Collective, Sept 1, 2026 (Circleback notes; internal)
  9. [9]Pricing, Financial Datasets
  10. [10]FactSet Meets Demand for AI-Ready Data, First to Announce MCP Sans Intermediary (December 16, 2025)
  11. [11]Kensho LLM-ready API FAQs
  12. [12]EDGAR Application Programming Interfaces (SEC)
  13. [13]Pricing, Massive (formerly Polygon.io)
  14. [14]Pricing Plans, Financial Modeling Prep
  15. [15]TauricResearch/TradingAgents data flows (GitHub)
  16. [16]Daloopa Raises $47 Million Series C (May 28, 2026)
  17. [17]S&P Global Launches Adaptive Retrieval (July 21, 2026)
  18. [18]What We Learned Building Enterprise MCP at Bloomberg
  19. [19]Morningstar and PitchBook Bring Trusted Investing Intelligence to Apps in ChatGPT (December 17, 2025)
  20. [20]FactSet (FDS) Q3 FY2026 earnings call transcript (July 1, 2026)
  21. [21]OpenAI launches ChatGPT Work (BNN Bloomberg, July 9, 2026)
  22. [22]Google Cloud Launches Gemini Enterprise for Financial Services (August 25, 2026)
  23. [23]Anthropic deepens finance push (Reuters via Yahoo Finance, May 2026)
  24. [24]JPMorgan enters the era of long-running autonomous agents (CNBC, June 9, 2026)
  25. [25]OpenAPI schema for api.financialdatasets.ai (GitHub gist)
  26. [26]virattt/ai-hedge-fund (GitHub)
  27. [27]virattt/dexter (GitHub)
  28. [28]Financial Datasets provider, Exa Connect docs
  29. [29]Financial Datasets vs Financial Modeling Prep (Find My Moat, 2026)
  30. [30]Bloomberg Terminal (Wikipedia)
  31. [31]Agent Adoption: The IT Industry's Next Great Inflection Point (IDC)
  32. [32]Intrinio Opens Institutional-Grade Financial Data to Everyone (June 2026)
  33. [33]Exa Raises $250M Series C (May 2026)
  34. [34]Exa Labs Raises $250M in Series C Funding at $2.2 Billion Valuation (FinSMEs, May 2026)
  35. [35]Exa API pricing (Search, Answer, Research, Monitors)
  36. [36]Parallel API Pricing
  37. [37]Parallel Web Systems hits $2B valuation five months after its last big raise (TechCrunch, April 29, 2026)
  38. [38]Introducing the Monitor API (Parallel)
  39. [39]Parallel for finance
  40. [40]Parag Agrawal's Parallel wants to pay publishers when AI agents use their work (Fortune, May 19, 2026)
  41. [41]Parallel Launches Index, a New Model for Compensating Content Owners (PR Newswire via Morningstar, May 19, 2026)
  42. [42]Training Data — Parallel's Parag Agrawal: Building a New Web for AI Agents (Sequoia, August 26, 2026)
  43. [43]firecrawl/firecrawl (GitHub)
  44. [44]Firecrawl: We just raised our Series A and shipped /v2 (August 2025)
  45. [45]Firecrawl pricing explained (2026)
  46. [46]7 Best Investment Research APIs for AI-First Use Cases in 2026 (Firecrawl)
  47. [47]Daloopa expands financial data MCP through a new connector with OpenAI
  48. [48]sec-api.io vs edgar.tools: Pricing & Free Tier Compared (2026)
  49. [49]Databento Raises $97 Million Series B Led by NEA (July 9, 2026)
  50. [50]anthropics/financial-services connector and plugin list (GitHub)
  51. [51]Introducing ChatGPT for Excel and new financial data integrations (OpenAI)
  52. [52]financial-datasets/mcp-server open issues (GitHub)