The SEC data your backtest actually needs.
Every other data vendor restates history, drops delisted companies, and patches filing dates after the fact. We don't. If it wasn't filed with the SEC on the date you query, it won't appear in your results.
- Bitemporal knowledge_at timestamps — no silent retroactive edits.
- Full universe including all delisted, bankrupt, and acquired companies.
- DuckDB-native Parquet: query 105M facts in milliseconds, locally.
- From pip install to production DataFrame in 60 seconds.
- MCP server for AI-accelerated research workflows.
from valuein_sdk import ValueinClient, ValueinError
try:
with ValueinClient() as client:
# PIT: what did the market know on 2020-01-01?
# Any restatement filed after that date is invisible.
df = client.query("""
SELECT fa.fiscal_year,
round(fa.numeric_value / 1e9, 2) AS revenue_bn,
fa.knowledge_at -- SEC acceptance timestamp
FROM fact fa
JOIN filing f ON fa.accession_id = f.accession_id
JOIN security s ON f.entity_id = s.entity_id
WHERE s.symbol = 'AAPL'
AND f.form_type = '10-K'
AND fa.standard_concept = 'TotalRevenue'
AND fa.knowledge_at <= '2020-01-01'
ORDER BY fa.fiscal_year DESC
""")
except ValueinError as e:
print(f"Error: {e}")
# fiscal_year revenue_bn knowledge_at
# 2019 260.17 2019-10-31T22:01Z ← last visible
# 2018 265.60 2018-11-05T06:01Z
# FY2020 10-K wasn't filed until Oct 2020 — invisibleWhy most financial datasets break your backtests.
You've seen it before: a factor that looks great in research falls apart live. Nine times out of ten, the culprit is the data — not the model.
Point-in-Time Accuracy
Every fact is tagged with knowledge_at — the exact EDGAR acceptance timestamp. Your 2017 backtest only sees data that was public in 2017. No silent restated imports.
Zero Survivorship Bias
Enron, Lehman Brothers, every bankrupt company that ever filed with the SEC is still here. Screen on the full population, not just today's survivors.
Native Parquet Format
Columnar Parquet files. DuckDB, Polars, and Spark read them natively. Query 105M rows locally for factor construction — zero egress on sample tier.
Sub-80ms API Latency
Cloudflare edge infrastructure. P50 under 80ms globally. P99 under 450ms. Your real-time trading model won't wait on your data provider.
Full Filing History
10-K, 10-Q, 8-K, 20-F — 30+ years, 12M+ filings. Amendment tracking built in: original reported vs. restated values, both available.
Schema Stability
Field names and types are versioned. Breaking changes never ship without a major version bump and a 90-day deprecation window. Your pipeline won't break on a Friday.
Real data. Right now.
AAPL 10-K annual fundamentals — standardized from SEC EDGAR (USD millions)
| ticker | period_end | revenue ($M) | net_income ($M) | eps_diluted | form_type |
|---|---|---|---|---|---|
| AAPL | 2024-09-28 | 391,035 | 93,736 | 6.11 | 10-K |
| AAPL | 2023-09-30 | 383,285 | 96,995 | 6.13 | 10-K |
| AAPL | 2022-09-24 | 394,328 | 99,803 | 6.11 | 10-K |
| AAPL | 2021-09-25 | 365,817 | 94,680 | 5.61 | 10-K |
| AAPL | 2020-09-26 | 274,515 | 57,411 | 3.28 | 10-K |
Sample tier. Revenue and net income in USD millions. Available via API, Python SDK, and MCP.
Add an AI layer to your quant workflow.
The Python SDK is for production pipelines. The MCP server is for the research phase — when you want to explore data, generate hypotheses, and validate signals fast, using Claude or Cursor as a research co-pilot.
Your agent queries the same underlying dataset as your SDK — same PIT accuracy, same full universe, same 105M facts. The difference is you interact in natural language instead of writing a query from scratch.
- →"What's the 5-year FCF CAGR for the S&P500 semiconductor sector?"
- →"Show me companies where accruals ratio spiked before a miss."
- →"Compare debt levels for regional banks pre and post 2020."
Based on NVIDIA's SEC 10-K filings — point-in-time · 2025-02-26
| Year | Revenue | YoY |
|---|---|---|
| FY 2024 | $60.9B | +122.1% |
| FY 2023 | $26.9B | +66.1% |
| FY 2022 | $16.7B | −0.2% |
| FY 2021 | $16.7B | +52.7% |
| FY 2020 | $10.9B | +52.9% |
valuein MCP · get-growth-metrics · NVDA · SEC EDGAR
Stop cleaning data. Start finding alpha.
Free S&P500 tier — no credit card, no demo call, no sales process. Sign up and your first DataFrame is two commands away.