Your research process, running on your entire universe — automaticallyYour research process, running on your entire universe — automatically. Your AI agent, now an equity research desk.
One connector, and the agent you already use reads SEC filings back to 1993 — Claude, Copilot, ChatGPT, Perplexity, Gemini, Grok, or Cursor. No new app, no migration. Every number it states traces to the filing it came from. mcp.valuein.biz/mcp
Select your AI — copy the config and connect in under 30 seconds.
Edit the config file, then restart Claude Desktop. Settings → Developer → Edit Config opens it directly.
Tip: Omit --headers to use the free sample tier.
%APPDATA%\Claude\claude_desktop_config.json{
"mcpServers": {
"valuein": {
"url": "https://mcp.valuein.biz/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN_HERE"
}
}
}
}- 111M+facts
- 19,000+companies
- 1993→history
- 118tools
- 39workflows
How it works
Point any MCP client — Claude, Copilot, ChatGPT, Gemini — at one endpoint. One token unlocks the same intelligence everywhere: point-in-time SEC datasets, 118 tools, 39 expert procedures, and your desk's persistent state — identical in every agentic platform you connect.
Valuein MCP Server
mcp.valuein.biz
point-in-time · survivorship-free
Agent-agnostic by design: your desk lives in the protocol — not in an app.
Same facts, same tools, same state — in Claude, Copilot, ChatGPT, and Cursor. Theses, watchlists, signals, and reports persist across sessions and across clients, keyed to your token. Save a thesis in Claude; list it from Cursor; score it a quarter later from whichever AI your firm runs that day.
Monday · in Claude
save_thesis> save_thesis { ticker: "NVDA", view: "long", horizon: 90d }
✓ saved · th_8f3a
Save the call where you did the work — in the AI you already use.
Tuesday · from Cursor
list_theses> list_theses
✓ th_8f3a · NVDA · long · 90d
A different app, the same desk. State is keyed to your token, not to a chat window.
Day 90 · any client
score_thesis_outcome> score_thesis_outcome { thesis_id: "th_8f3a" }
✓ graded vs filings + price · track record updated
The desk remembers what you said a quarter ago — and grades it. Your agent keeps score.
Your desk
keyed to your token
- thesis
th_8f3a · NVDA · long - watchlist
AI infra · 12 names - signal
fundamental_change · armed - report
NVDA brief · v3
No export. No sync.
No per-seat app.
Your agent, our facts — the same desk in every AI.
One connector. 118 tools. 39 expert procedures.
Typed, tier-gated tools — not a chat wrapper. One token unlocks the identical surface in every MCP client you connect, and every time-series tool accepts an as_of_date, so the same call reconstructs what was knowable on any historical day.
Standardized fundamentals, filings, and one-click lineage.
get_company_fundamentalsget_financial_ratiosget_valuation_metricsverify_fact_lineage
Survivorship-free universes and cross-sectional screens.
screen_universeget_pit_universeget_peer_comparablesget_earnings_signals
Insider Forms 3/4/5/144 and 13F/13D/13G ownership flows.
get_insider_transactionsget_institutional_holdingsget_smart_money_flowget_top_holders
The memory layer: persists across sessions and clients.
save_thesissave_watchlistcreate_signalschedule_taskcreate_rule
Deterministic valuation and audit work, done server-side.
compute_dcfforensic_auditrun_backtest
Branded, board-ready artifacts straight from the desk.
generate_dcf_xlsxgenerate_research_brief_docxgenerate_comps_xlsx
The 39 procedures are expert playbooks your agent runs end-to-end — flagships equity_research_brief and screen_and_shortlist chain the tools into a complete, reproducible brief. See the workflows below ↓
Start free. Scale when you need more.
S&P500 data is free — no credit card. Your Bearer token determines data scope and rate limits, enforced at the edge on every call.
S&P500 · 500+ tickers · 1993–present
60 req/min · 1,000 req/hr
Active + delisted US universe · 19,000+ entities · 15-year history (2011→present)
100 req/min · 3,000 req/hr
Two datasets: fundamentals (111M+ facts, 19,000+ entities, 1993–present) + smart-money (~78M rows · 6 tables: Forms 3/4/5/144 + 13F/13D/13G)
300 req/min · 10,000 req/hr
| Tier | Data scope | Rate limit | Price |
|---|---|---|---|
S&P500Free | S&P500 · 500+ tickers · 1993–present | 60 req/min · 1,000 req/hr | Free |
ProPopular | Active + delisted US universe · 19,000+ entities · 15-year history (2011→present) | 100 req/min · 3,000 req/hr | $49/mo |
Institutional | Two datasets: fundamentals (111M+ facts, 19,000+ entities, 1993–present) + smart-money (~78M rows · 6 tables: Forms 3/4/5/144 + 13F/13D/13G) | 300 req/min · 10,000 req/hr | $499/mo |
S&P500 tier — free, no credit card required.
39 expert procedures — or codify your own.
Tools return data. Procedures return analysis. Each one is an ordered chain of typed tool calls — the exact sequence a senior analyst, quant, or PM would run — executed the same way every time. Your process, codified: same inputs, same output, on every name you cover.
equity_research_briefFinancial AnalystSingle-ticker end-to-end institutional research brief in markdown — fundamentals, valuation, ratios, capital allocation, peer comparison, recent catalysts, and SEC lineage. Three depth modes: 'quick' (≈3 tool calls — snapshot), 'full' (≈8 calls — default institutional brief), 'forensic' (≈11 calls — adds restatement audit + fact-level SEC verification). Renders as an artifact users can export to Word/PDF directly from Claude Desktop or claude.ai. PIT-safe via as_of_date for backtests.
search_companiesget_company_fundamentalsget_valuation_metricsget_financial_ratiosget_capital_allocation_profileget_peer_comparablesget_earnings_signalsget_sec_filing_linksverify_fact_lineagescreen_and_shortlistPortfolio ManagerPM idea-generation workflow. Builds a survivorship-free universe via index_membership.parquet, ranks it on a chosen factor objective (quality, value, growth, or balanced), QCs the top 3 with a period-over-period change check, and hands off shortlist members to equity_research_brief for full write-ups. Renders as a markdown shortlist artifact. Survivorship-free historical screening via as_of_date.
get_pit_universescreen_universecompare_periodsequity_research_briefmargin_and_moat_teardownFinancial AnalystSystematic 5-year analysis of a company's operational efficiency and competitive moat. Chains fundamentals → valuation metrics → peer comparables to produce a structured teardown covering revenue quality, margin trends, ROIC vs WACC spread, and relative positioning vs sector peers.
get_company_fundamentalsget_valuation_metricsget_financial_ratiosget_peer_comparablespeer_benchmarking_memoFinancial AnalystProduces a structured relative-value memo comparing a company to its closest peers. Chains search_companies → peer_comparables → valuation_metrics → financial_ratios to output an investment-committee-ready comparison table with ROIC, valuation multiples, and capital efficiency.
search_companiesget_peer_comparablesget_valuation_metricsget_financial_ratiosquality_and_risk_auditPortfolio ManagerEvaluates capital structure safety and dividend sustainability for portfolio risk-adjusted sizing. Chains fundamentals → capital_allocation → valuation_metrics → financial_ratios (leverage/liquidity) to produce a structured risk scorecard.
get_company_fundamentalsget_capital_allocation_profileget_valuation_metricsget_financial_ratiosyour_own_procedureYouWrite the steps in the exact order your process demands. Your edge becomes a procedure your agents run identically on every name — not a one-off deck.
Run it across your whole universe — deterministically.
Point your agents at your sector, your index, or your circle of competence and run the same workflow across every name at once. Same inputs, same output, every time — every number pinned to its filing.
run_workflow
1 · get_company_fundamentals
2 · get_financial_ratios
3 · get_peer_comparables
4 · compute_dcf
The steps you chose, in the order you chose them — your process, codified.
… and every other name you cover, out of 19,000+ back to 1993
Your agent team runs the same procedure on each ticker, in one batch — no one runs it by hand twelve times.
- NVDAbrief v5 · updated Thu
- JPMbrief v2 · updated Fri
- LLYbrief v3 · updated Fri
Every run lands as a persisted, versioned report — a library, not a pile of chats.
Not “an AI opinion.” A reproducible analysis you can defend.
Keep it current — automatically.
Set an automation once. When a new 10-K, 10-Q, or 8-K changes the numbers, your team of AI agents re-runs the analysis and updates the profile — so the firm is always working from today's filings, not last quarter's.
A filing lands
A 10-K, 10-Q, or 8-K is accepted at EDGAR and ingested with its acceptance timestamp — history stays append-only.
accepted_at stampedThe rule fires
The automation you set once matches the new numbers and triggers the workflow — no one has to notice first.
create_rule · fundamental_changeYour agent team re-runs the analysis
The identical workflow executes again — same steps, same order, temperature 0, every figure re-pinned to the new filing.
run_workflow · temperature 0An updated profile + a notice
The company profile gets a new version and an inbox item tells you what changed and which filing changed it.
report v2 · inbox: run_completedschedule_taskWhen the 10-K lands, your agents re-run the work. The firm wakes up current.
Anything that mutates state or reaches the outside world stages for your approval first — human-on-the-loop by default, an immutable audit entry every time.
One analyst's research. Every desk's source of truth.
A research library so rigorous your credit desk can price a loan from it, your M&A team can scope a deal with it, a trader can check a risk before sizing — and your compliance team can sign off on it — because every number traces to the filing.
Price a loan off the same risk read — leverage and coverage straight from the filings.
Scope a target from a valuation everyone can trace to source.
Check a company's risk before sizing a position.
Rebalance on company reads that refresh with every new filing.
Sign off with confidence — every number traces to the filing.
One source of truth — not twelve conflicting decks.
Why not Bloomberg, WRDS, or Nasdaq Data Link?
Same SEC EDGAR source data. Radically different access model. Valuein is the only one with a native MCP surface, open-source SDK, and self-serve pricing that starts at zero.
| Dimension | Valuein | Bloomberg | WRDS / Compustat | Nasdaq Data Link |
|---|---|---|---|---|
| Starting price | Free · $49 · $499/mo | ~$20K+/yr per seat | Institutional license | Per-dataset, adds up |
| MCP server for AI agents | 118 tools + 39 playbooks, streamable HTTP | — | — | — |
| Persistent agent state | Theses, signals, schedules — same state in every AI | — | — | — |
| Python SDK | Open source on PyPI | License-gated API | Proprietary library | Proprietary |
| Point-in-time | accepted_at on every fact | BCORP license required | Snapshot-based | Varies by dataset |
| Survivorship bias | Zero — delisteds included | Config-dependent | Compustat universe | Varies |
| Bulk Parquet export | Presigned URLs, DuckDB-ready | BSYM extract | SAS / CSV | CSV / JSON |
| Self-serve signup | Active in minutes | Sales contract | Institutional affiliation | Marketplace checkout |
Competitor pricing reflects publicly reported list ranges, not quotes. Full side-by-side breakdowns: vs Bloomberg · vs WRDS · vs Nasdaq Data Link.
The model never mints a number.
Every figure is born in a filing and carries its fact_id. Every numeric response ships a lineage envelope — the fact_id, the source filing, and a URL an auditor can open. The round-trip is a tool call, not a research project.
A figure lands in the answer
Your agent states Apple's FY2024 revenue. The number arrived from a tool — typed, with provenance attached — never from the model's memory.
revenue: $391,035,000,000 fact_id: "fct_0000320193_FY2024_Revenues_9e4a"
One call resolves its lineage
verify_fact_lineage takes the fact_id and returns the source accession number, both filing timestamps, and whether the value was ever restated.
verify_fact_lineage({ fact_id: "fct_…9e4a" })
→ 0000320193-24-000123 · 10-K · restated: noThe exact filing, on sec.gov
The response carries a direct EDGAR URL. Compliance opens the 10-K the number came from and reads it at the source — no Valuein login required.
→ sec.gov/Archives/edgar/data/320193/…-index.htm
An auditor doesn't re-run anything. They click.
Verifiable facts
Every figure carries a fact_id traceable to the exact SEC filing it came from. Numbers are born in the data — the model never invents one.
Reproducible runs
Deterministic, typed tools; managed agent runs at temperature 0; point-in-time snapshots. Same inputs, same output — research your firm can re-derive later.
Governance is shipped, not promised.
High-impact actions stage for a human's approval before they run, risk-tiered and logged to an immutable audit trail. Live today — not a roadmap item.
Built for agents, secured like infrastructure.
Financial data buyers don't ship agents against a black box. Here's exactly what the MCP surface can and can't do.
Bearer token auth, edge-validated
Every request is validated at the nearest edge. Tokens are tier-scoped and revocable in one step — no shared secrets, no long-lived API keys in client config.
No execution surface
The data surface is read-only SEC filings; the only thing your agent can write is your own research state — theses, watchlists, signals, reports. No order tools, no fund transfers, no broker access. Ever.
No PII on the data surface
We store token hashes and tier metadata. The data your agent queries is public SEC EDGAR filings. Nothing user-identifying touches the MCP response path.
Edge-enforced rate limits
Per-token limits run at the edge (60 / 100 / 300 req/min by tier). If your agent loops, we throttle — a clean 429, not a runaway bill.
Human-on-the-loop approvals
Read-only tools run free. Anything high-impact stages first — risk-tiered, waiting on a human's explicit approval, and recorded in an immutable audit log.
Open protocol, no lock-in
Streamable HTTP MCP is a public spec. Swap clients, swap agents, self-host the reader — your token keeps working, the wire format stays the same.
Your agent can onboard itself.
The server's discovery surface is written so an MCP-capable agent can inspect the tool schema, hit the free tier, and — if it needs more scope — point its human operator at checkout. No sales call required.
Frequently asked questions
Add one connector.
Get a research floor.
Start with the free tier — real filings in under 15 seconds, no card. Pro full-universe + 15-year history at $49/mo. Institutional with smart-money data + webhooks + redistribution at $499/mo. One token, the same intelligence in every AI you connect.