Finsyt

The data plane, then the research workflow.

Finsyt is one MCP and one REST API for entitled market data, filings, and macro — with history and provenance — plus a desk that turns those figures into memos, models, and citations.

One plane. Two transports.

Agents speak MCP. Services speak REST. Both hit the same handlers, the same entitlements, and the same provenance envelope. FQL mnemonics (FX_REV, FX_EBITDA) are the public language — never vendor shorthand.

  • Hosted MCP at /api/v1/mcp — Cursor, Claude, any streamable-HTTP client.
  • REST /api/v1 for notebooks, internal apps, and Excel.
  • Catalog, candles, series, options, filings, news, macro — one key.
  • MCP Apps visualize financials, prices, valuation, and deals inside the host. Skills cover tearsheets, earnings, DCF, and Excel FQL.
Developers Coverage
GET /api/v1/financials
{
  "symbol": "AAPL",
  "metric": "FX_REV",
  "value": 416161000000,
  "source": "sec",
  "citation": {
    "tag": "us-gaap:RevenueFromContractWithCustomerExcludingAssessedTax",
    "accession": "0000320193-25-000079",
    "filingDate": "2025-10-31",
    "filingUrl": "https://www.sec.gov/Archives/..."
  },
  "provenance": {
    "dataset": "sec_xbrl",
    "as_of": "2025-09-27",
    "retrieval_method": "sec_edgar"
  }
}

Semantic precision, not keyword matching.

Ask complex financial questions exactly how you'd ask an analyst. Finsyt understands context, synonyms, and financial jargon, pulling the exact right data points from thousands of documents simultaneously.

  • "Summarize the key drivers of margin compression for AAPL in Q3 vs Q2."
  • "Create a table of restructuring charges for top 5 banks over the last 3 years."
You
What is NVDA's capital return strategy?
F

NVIDIA's capital return strategy is primarily focused on share repurchases and a modest dividend. In FY2024, the company returned $9.9 billion to shareholders in the form of share repurchases and cash dividends.

NVDA FY24 10-K, p. 38

Extraction Table

MetricQ1 '24Q2 '24
Revenue$26.0B$30.0B
Gross Margin78.4%75.1%
OpEx$3.5B$3.9B

Structured outputs, ready for your model.

Don't just read answers—use them. Finsyt automatically extracts complex financial tables, normalizes data across different companies, and allows one-click export directly into your Excel models.

Every answer shows its work.

Compliance teams shouldn't have to take an LLM's word for it. Every Finsyt answer — in Research, AI Analysis, the Decision Matrix, and exported memo decks — ships with a "Data sources used" footer that lists each provider and connector that fed the response, the role it played, the time it took to respond, and a deep link straight to the Connector Hub.

  • Provider, role, response time, citations — surfaced beside every answer.
  • Deck appendix slide mirrors the chat footer so investment memos stay self-attributed.
  • Workspace-wide toggle with per-analyst collapse — defaults on for audit-ready output.
Data sources used
  • PrimaryFinancial Modeling Prep
    421 ms Hub
  • FallbackEOD Historical Data
    612 ms Hub
  • CitationSEC EDGAR · 10-K
    3 cites Hub
Hosted MCP serverLive
$ claude mcp add --transport http finsyt \
    https://finsyt.com/platform/api/v1/mcp \
    --header "Authorization: Bearer fsk_…"

> finsyt connected · 30+ tools
> finsyt_workspaces_list
> finsyt_notes_list · finsyt_note_create
> finsyt_peer_compare · finsyt_alerts_list

Bring your own AI agent.

Connect Claude, Cursor, or any MCP-compatible agent with a single API key. Your agent gets the Finsyt catalog — market history, filings, notes, peers, alerts — tenant-scoped, entitled, and audited. Never anyone else's data.

  • Scoped keys — read-only by default, write access and per-key tool allowlists when you need them.
  • Full audit trail — every agent tool call is logged to your workspace's audit history.
  • Works everywhere — one endpoint for Claude Code, Claude Desktop, Cursor, and any streamable-HTTP MCP client.
Connect your agent in the developer console