GateKey gives your agent schema-stable JSON from SEC filings, earnings calls, market data, and research papers - with provenance, token budgets, and cross-provider joins. One API key. One request shape.
npx @gatekey/mcp-finance(soon)// One shape. Every provider. POST https://api.gatekey.dev/v1/query { "provider": "sec-edgar", "endpoint": "/company-facts", "params": { "ticker": "TSLA", "facts": ["Revenue", "NetIncome"], "periods": 8 }, "controls": { "max_tokens": 2000, "include_provenance": true } }
{
"data": {
"ticker": "TSLA",
"facts": [{"metric": "Revenue",
"value": 25_500_000_000,
"period": "2024-Q3"}]
},
"provenance": {
"accession": "0001318605-24-015",
"filed": "2024-10-23",
"source_url": "https://sec.gov/..."
},
"meta": {
"tokens_used": 847,
"latency_ms": 312,
"budget_remaining": 99_153
}
}Web scrapers (Firecrawl, Cloudflare Markdown) convert HTML pages into text. GateKey queries specialized data sources and returns structured, schema-stable JSON with provenance, token counts, and bounded outputs.
Scrape a web page - get unstructured text. No schema guarantees. No provenance. Token count varies wildly. Output changes when the page layout changes.
firecrawl.scrape("sec.gov/cgi-bin/browse-edgar?...")# Tesla Inc - 10-K Annual Report
Filed: October 23, 2024
## PART I
### Item 1. Business
We design, develop, manufacture, sell and
lease high-performance fully electric vehicles
and energy generation and storage systems...
# 16,000+ tokens of raw text
# No structured financials
# No accession number
# Schema changes on next scrapeQuery a data source - get normalized JSON with stable schema, provenance pointers, and exact token budget. Output is consistent across every company, every filing.
gatekey.query("sec-edgar", "/company-facts", ...){
"ticker": "TSLA",
"facts": [
{
"metric": "Revenue",
"value": 25500000000,
"period": "2024-Q3",
"unit": "USD",
"xbrl_tag": "us-gaap:Revenues"
}
}],
"provenance": {
"accession": "0001318605-24-015",
"source_url": "https://sec.gov/...",
"section": "Financial Statements"
},
"meta": { "tokens_used": 847 }
}Every response from GateKey is engineered for AI consumption - predictable, bounded, traceable.
Every provider returns the same JSON shape. Tesla revenue looks identical to Apple. Your agent code never breaks when we add providers or sources update their formats.
Set max_tokens per request. GateKey shapes the response to fit your context window - no truncation surprises, no blown budgets. Every response includes exact token count.
Every fact links to its source: accession numbers, filing URLs, section IDs, publication dates. Your agent can cite its work. Auditors can verify it.
One query can pull SEC filings + earnings transcripts + news. GateKey joins across providers so your agent does not need to orchestrate five separate API calls.
Six financial data providers, structured for agents. One MCP install gives your local Claude or Cursor access to SEC filings, earnings calls, insider trades, market data, and macro indicators.
Not raw XBRL. Not adversarial HTML. Normalized financial tables, consistent JSON schemas across 10,000+ companies, section-level extraction from 10-Ks, 10-Qs, and 8-Ks.
Full earnings call transcripts chunked for context windows. Management sentiment extraction, exact CEO/CFO quote attribution, Q&A section parsing. Via Financial Modeling Prep.
Real-time SEC Form 4s + congressional stock transactions. Parsed, enriched with committee memberships and disclosure dates. Powered by OpenInsider + Capitol Trades.
Federal Reserve Economic Data: CPI, employment, housing, GDP, interest rates. Cleaned time-series JSON, ready for agent analysis. No more archaic XML transformation.
Live stock prices, options chains with Greeks, and fundamentals via Polygon.io / Alpha Vantage. Synchronous REST for agents - no WebSocket complexity.
150K+ sources via NewsAPI. Filtered by ticker, sector, or topic. Token-bounded summaries with sentiment scores and source attribution.
import requests, feedparser, xml.etree
from alpha_vantage.timeseries import TimeSeries
from fredapi import Fred
# SEC EDGAR - parse XBRL yourself
headers = {"User-Agent": "me@email.com"}
r = requests.get(f"https://efts.sec.gov/...",
headers=headers)
# handle rate limits (10 rps)
# parse nested XML/HTML
# normalize financial tables manually
# extract accession numbers manually
# Alpha Vantage - different auth, schema
ts = TimeSeries(key='YOUR_KEY')
data, meta = ts.get_quarterly('TSLA')
# 5 calls/min on free tier
# different JSON shape than EDGAR
# FRED - yet another auth + format
fred = Fred(api_key='ANOTHER_KEY')
cpi = fred.get_series('CPIAUCSL')
# returns pandas Series, not JSON
# NewsAPI - yet another key
news = requests.get("https://newsapi.org/v2/...",
params={"apiKey": "THIRD_KEY", ...})
# Now join all of these together...
# Handle 4 different error formats...
# Count tokens for your context window...
# Hope nothing breaks tomorrow.from gatekey import GateKey gk = GateKey(api_key="gk_live_...") # SEC + Earnings + News in one call result = gk.query( provider="sec-edgar", endpoint="/company-facts", params={ "ticker": "TSLA", "facts": ["Revenue", "NetIncome"], "periods": 8 }, controls={ "max_tokens": 2000, "include_provenance": True } ) # Stable schema. Every time. revenue = result.data.facts[0].value source = result.provenance.accession tokens = result.meta.tokens_used budget = result.meta.budget_remaining # Same shape for FRED, News, Earnings macro = gk.query("fred", "/series", params={"series_id": "CPIAUCSL"}) # Works as a LangChain tool tool = gk.as_langchain_tool()
Sign up, get a key, query any provider. Same request shape for everything.
One command for MCP, SDK, or raw REST. Works with Claude, Cursor, LangChain, CrewAI, and any agent framework.
pip install gatekey-ai (soon)npx @gatekey/mcp-finance (soon)One request shape for every provider. POST /v1/query with provider, endpoint, params, and controls. Get structured JSON back.
Token budgets prevent surprises. Provenance metadata enables citations. Scoped keys keep production safe. Agent-ready from day one.
As MCP ecosystems grow, security is not optional. GateKey ships with the guardrails production agents need from day one.
Restrict keys to specific providers and endpoints. Your SEC-only key cannot query PubMed. Your read key cannot trigger writes.
Set daily or monthly hard caps per key. When the budget hits zero, the key stops - no runaway costs from autonomous agents.
Every query logged with timestamp, provider, token count, and response hash. Export for compliance reviews or SOC 2 evidence.
max_tokens, include_fields, top_k, dedupe - control exactly what comes back. No context window blowouts.
We are building toward a marketplace where data providers set prices per token, and agents pay autonomously - no enterprise contracts, no manual billing. Powered by x402 agent payment rails.
Providers list content and set royalties. Agents discover, query, and pay - by the token. Developers surface what data they would pay for. The settlement layer between AI agents and the world data.
// Coming: agent-initiated data purchases POST /v1/marketplace/query { "provider": "bloomberg-news", "endpoint": "/article", "params": { "topic": "TSLA earnings" }, "payment": { "protocol": "x402", "max_spend": "0.05 USDC", "wallet": "agent_wallet_0x..." } } // Response includes payment receipt { "data": { ... }, "payment_receipt": { "tx_hash": "0xabc...", "amount": "0.012 USDC", "provider_royalty": "0.008", "gatekey_fee": "0.004" } }
Start free. Upgrade when your agent needs more data. Every plan includes all agent-safe controls.
We have built data infrastructure for frontier AI labs and Fortune 500 operations floors. We know what it takes to make data production-grade.
Previously built data infrastructure at Scale AI (serving frontier AI labs, growing from 3 to 50+ enterprise customers) and led operations at McKinsey. We have seen firsthand how painful it is to connect AI agents to real-world data sources - and we are building the infrastructure we wished existed.
250,000 free tokens. SEC filings, earnings calls, macro data. First query in under 3 minutes.