Introducing EarningsBench: we tested 6 LLMs on SEC filing extraction. [Learn more.](https://www.financialdatasets.ai/blogs/earningsbench)

FINANCIAL DATASETS
Products:
  [Financial Statements](https://docs.financialdatasets.ai/api/financials/income-statements)
  [Stock Prices](https://docs.financialdatasets.ai/api/prices/historical)
  [SEC Filings](https://docs.financialdatasets.ai/api/filings/ticker)
  [Insider Trades](https://docs.financialdatasets.ai/api/insider-trades/insider-trades)
  [Institutional Holdings](https://docs.financialdatasets.ai/api/institutional-holdings/ticker)
  [Earnings](https://docs.financialdatasets.ai/api/earnings/earnings)
Developers:
  [Docs](https://docs.financialdatasets.ai/introduction)
  [Quickstart](https://docs.financialdatasets.ai/quickstart)
  [MCP Server](https://docs.financialdatasets.ai/mcp-server)
  [agents.md](https://www.financialdatasets.ai/agents.md)
  [OpenAPI Spec](https://www.financialdatasets.ai/openapi.json)
  [llms.txt](https://www.financialdatasets.ai/llms.txt)
  [Support](mailto:support@financialdatasets.ai)
Resources:
  [Blog](https://www.financialdatasets.ai/blogs)
  [Pricing](https://www.financialdatasets.ai/pricing)
  [Enterprise](https://www.financialdatasets.ai/enterprise)
  [Contact](https://www.financialdatasets.ai/contact)
[LOG IN](https://www.financialdatasets.ai/login)
[SIGN UP](https://www.financialdatasets.ai/register)

# Financial data at scale

Built for investors and AI teams.

## The default data layer for your agents
Clean, point-in-time data, auditable to the source. Seconds after the report.
The stack, top to bottom:
- Your firm: The theses, models, and memos your PMs act on, each number traceable to its source.
- Your agents: Agents that run long-horizon research tasks, from screens to memos.
- Financial Datasets: Accurate, auditable data for AI, and the largest financial search index for agents.
- Raw primary sources: Filings, earnings releases, transcripts, and prospectuses, each in its own format.


## We clean the data. You find the alpha.
We map complex, disparate data across global markets into clean rows.
Example run: How did Nvidia's growth mix shift in Q2 FY27?
- get_earnings(NVDA) -> revenue $96.2B, beat 15.0%
- get_kpi_metrics(NVDA, Q2 FY27) -> Data Center $89.0B
- get_kpi_metrics(NVDA, Q2 FY26) -> Hyperscale $24.2B, ACIE $16.9B
- get_kpi_guidance(NVDA) -> Q3 revenue $108.0B ±2%
- get_kpi_non_gaap(NVDA) -> gross margin 75.0%
Revenue reached $96.2B, up 106% y/y and 15.0% above estimates. [[1] NVDA 8-K, EX-99.1](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27pr.htm#:~:text=Revenue%20of%20%2496.2%20billion%2C%20up%20106%25%20from%20a%20year%20ago)
AI Clouds & Enterprise grew 138% y/y, ahead of Hyperscale at 102%. [[2] NVDA CFO commentary](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27cfocommentary.htm#:~:text=AI%20Clouds%2C%20Industrial%2C%20%26%20Enterprise,40%2C313)
Q3 guide is $108.0B ±2%, with no China data center revenue assumed. [[3] NVDA 8-K, EX-99.1](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000073/q2fy27pr.htm#:~:text=Revenue%20is%20expected%20to%20be,China%20in%20its%20outlook)

## Where your agents run
Financial Datasets plugs into Claude, ChatGPT, Hermes, or your own harness.
- Claude: claude.ai, Desktop, Claude Code
- ChatGPT: plugin, Codex
- Hermes Agent: MCP
- Custom harness: MCP, REST, bulk delivery

## For investment teams
Point-in-time history, as the market saw it at the time.
- 30 years of point-in-time financials, restatements included.
- Delisted companies stay in fundamentals: no survivorship bias.
- Every number links to the sentence it came from.
- Zero data retention: your research queries are never stored.

## For AI companies
Data your product can show its users, with redistribution in the contract.
- Request volume set to your needs, on dedicated compute.
- Licensing written for AI products: redistribution included.
- Webhooks that push new data the moment it lands.
- Uptime SLA written into your contract.

## Frequently asked questions

How is Enterprise priced?
Pricing is based on request volume, bulk datasets, and the terms you need. Tell us your expected volume and the datasets you use, and we send a quote after one call. Enterprise is billed by invoice with net terms.

How is bulk data delivered?
Parquet files in your S3 bucket by default, split by dataset and year, with a manifest that lists row counts and checksums. Snowflake share, Databricks Delta share, and CSV over SFTP are also available. Every financial row carries its filing date and as-of date, so point-in-time stays intact after the load.

What does zero data retention mean?
Your queries and responses are never retained. No logs, no audit trails, no persistence on our side.

Can we show the data to our own users?
Yes. Professional and Enterprise plans include data redistribution rights, with attribution to Financial Datasets. This covers figures, tables, and charts in your product, and data passed to an AI model that answers your users. [Terms of use](/terms-of-use)

How does a security review work?
Request the MSA, DPA, and security policies through the contact form. Our SOC 2 Type I audit is in progress. The subprocessor list and privacy policy are public. [Contact sales](/contact)

[CONTACT SALES](https://www.financialdatasets.ai/contact)
[PRICING](https://www.financialdatasets.ai/pricing)

Latest updates

June 29
[WE TESTED 6 LLMS ON SEC FILING EXTRACTION](https://www.financialdatasets.ai/blogs/earningsbench)

March 26
[ANNOUNCING OUR PARTNERSHIP WITH JULIUS](https://www.financialdatasets.ai/blogs/julius-ai-partnership)

CONTACT
  [support@financialdatasets.ai](mailto:support@financialdatasets.ai)

LEGAL
  [Terms of Service](https://www.financialdatasets.ai/terms-of-use)
  [Privacy Policy](https://www.financialdatasets.ai/privacy-policy)
  [Subprocessors](https://www.financialdatasets.ai/subprocessors)

[X / TWITTER](https://x.com/findatasets)

Financial Datasets, Inc. 2026