Stop writing the EDGAR parser. Real data in 15 seconds.
Skip the EDGAR parser. pip install, pull standardized Parquet with no credit card, and give your AI agent typed tools with fact-level lineage — plus state that persists across sessions and clients.
- pip install → real data in under 15 seconds, no API key on the sample tier.
- One normalized API across every filer — never map a raw XBRL tag again.
- 99 typed, tier-gated, deterministic MCP tools — plus SOP prompts that chain them into full workflows.
- State that persists across AI clients — save a thesis in Claude, read it in Cursor.
Built for
Developers
- Point-in-time accurate
- Survivorship-bias-free
- Every number cited to its filing
Works where you do
The pain points we remove
Technical buyers don't want a pitch — they want the thing they spent last weekend fighting to be already solved. Here's what we take off your plate.
Everyone rewrites the EDGAR parser
XBRL is technically public and free, but unusable raw. Devs burn 80% of their time wrestling EDGAR instead of building the actual product.
One concept, six tags
Revenue alone is us-gaap:Revenues, SalesRevenueNet, and RevenueFromContractWithCustomerExcludingAssessedTax — and different filers pick differently.
EDGAR's rate limits make apps fragile
10 req/sec per IP, hard 429s, ~10-minute IP blocks, a mandatory User-Agent — you have to build a whole caching/queue layer before you ship a feature.
Commercial APIs are pricey, throttled, and fragile
Free tiers are too thin to develop on, paid jumps are steep — and IEX Cloud shutting down took thousands of apps dark overnight. Vendor durability is now a real risk.
AI agents need real tools, not scraping
MCP is now the default way agents reach data — 78% of enterprise AI teams already run MCP-backed agents in production. But a 2026 security sweep found 43% of public MCP servers carry a known vulnerability. An agent that ships a made-up figure, or calls a poisoned tool, is a support ticket at best. Ours is typed, deterministic, and tier-gated from day one.
The grind we take off your plate
From the daily check-ins to the month-end scramble — this is the recurring work Valuein automates so you spend your hours on the thesis, not the data.
Every day
- Build features against the data API
- Debug auth, rate-limit, and normalization edge cases
- Handle the one filer whose tags don't match
Every week
- Add coverage — new tickers and statements
- Write retry/backoff and caching glue
- Reconcile data discrepancies users report
Month-end & earnings
- Watch the new filing wave for parser breakage
- Cost review as API usage scales
- Re-evaluate vendor risk after any provider outage
What you can do with Valuein
Each job you need done, mapped to the exact capability that delivers it.
pip install → real data, no credit card
The sample tier streams real Parquet with no API key, so you can evaluate before you commit.
One normalized API across all filers
Canonical concepts mean one Revenue, one EBIT — never map a raw XBRL tag again.
A drop-in toolset for your agent
99 typed tools in Claude, Copilot, ChatGPT, or Cursor over Streamable HTTP, behind one Bearer token — every figure returned with its fact_id.
Give your agent memory and a schedule
Theses, watchlists, alerts, and reports persist server-side across sessions and clients; agents can schedule deferred tasks and set rules that re-run work when a filing lands.
Skip the rate-limit/caching layer
The edge-gateway streams Parquet with tier-based limits — no EDGAR throttling to engineer around.
A vendor that won't disappear
Open Parquet and an open-source SDK on PyPI — your data and pipeline are portable, not locked in.
Works where you do
One Bearer token reaches the same point-in-time data from your AI agent, your notebook, or your browser. Use the surface that fits the job.
Stop writing the EDGAR parser. pip install valuein-sdk → real Parquet in 15 seconds, no credit card.
One canonical Revenue across 19,000+ filers. We mapped the 11,966 raw XBRL tags so you don't have to.
99 financial-data tools your AI agent can call over MCP — typed, gated, deterministic. Not HTML scraping.
Frequently asked
Can I really try it with no credit card?
Yes. pip install valuein-sdk and the sample tier (S&P 500, last 5 years) streams real Parquet with no signup, no email, no API key. Upgrade to Pro for the full 19,000+ universe and 15-year history.
How do I give my AI agent access to the data?
Add https://mcp.valuein.biz/mcp as a custom MCP server in Claude Desktop, Cursor, or any MCP client. It exposes 99 typed, tier-gated tools behind a Bearer token — the same token that unlocks the SDK.
Can my agent keep state between sessions?
Yes. Theses, claims, watchlists, alerts, and reports persist server-side, keyed to the token — save something from Claude and read it back from Cursor. Agents can also schedule deferred follow-ups and register rules that fire when an alert or filing event lands.
Will I get rate-limited like I do on EDGAR directly?
No. The edge-gateway enforces generous, tier-based limits (100 req/min on Pro, 300 req/min on Institutional) and streams pre-built Parquet, so you don't build a caching/backoff layer just to stay under SEC's 10 req/sec cap.
What happens to my integration if Valuein changes the schema?
The Parquet schema is versioned and manifest-driven — the SDK and MCP read it at runtime, and breaking changes ship behind a major version bump. The data is open Parquet, so you're never locked into a proprietary blob.
Stop writing the EDGAR parser. pip install valuein-sdk → real Parquet in 15 seconds, no credit card.
111M+ standardized SEC facts across 19,000+ companies, 1993–present. Free to start — no credit card.