Fiscal.ai

Fiscal.ai — the platform that traded as FinChat until 17 June 2025 — is an AI research terminal for public-company financials. The rename came alongside a $10M Series A and a stated ambition to be an operating system for finance rather than a chat window over filings (stockmarketstack.com's product profile; Fiscal.ai's own rebrand announcement, which also dates the original FinChat launch to 2023). The core idea is unchanged from the FinChat days: structured, standardized financial data for well over 100,000 stocks, wrapped in a conversational copilot that answers questions against that data and — critically — links every number back to the filing it came from. In my testing, that last part is what separates Fiscal.ai from the wave of generic finance chatbots: the product is built around verifiability, not around fluent answers.

What it actually does

There are two products, priced separately. The first is the web terminal: standardized and as-reported financials, segment and KPI breakdowns, earnings transcripts, SEC filings, insider and 13F ownership data, and fund letters. You get charting, screening, portfolio tracking, news, alerts, and an AI Copilot that answers questions in natural language against the dataset. The second is the same database sold as infrastructure: a REST API with webhooks, and an MCP server at api.fiscal.ai/mcp that plugs Fiscal.ai into Claude Desktop, Cursor, VS Code Copilot and similar tools as a live data source. There is an official Claude connector; other clients authenticate by OAuth or bearer key, and webhooks fire on data updates rather than forcing polling.

The pricing stack, at time of writing (fiscal.ai pricing as summarized in 2026 breakdowns and the stockmarketstack profile): Free at $0, Pro at $39/month billed annually or $49 month-to-month, and Max at $79/month billed annually or $99 month-to-month. The free tier gets ten years of annual financials and six quarters of recent periods, two years of KPI and segment data, one dashboard capped at 30 rows, and basic summaries. Pro lifts history to fifteen years and ten years of KPIs and adds cross-document analysis and estimate revisions. Max is where the two differentiating features live: click-through audit to filing, and financial-data export to Excel and CSV. Team and institutional deals are quoted; the API/MCP subscription publishes no price and bills annually — third-party breakdowns put small-team API access in the low hundreds of dollars per month, but I could not confirm that against a vendor page, so treat it as an estimate.

One more piece of the product worth knowing: brokerage connections. You can link holdings into your dashboard — Interactive Brokers and Kraken each have published setup guides, with IBKR data arriving in near real time. This is portfolio context for your research, not a trading integration.

My testing notes

First observation: the Copilot's answers are grounded in a way most finance chat tools are not. Asking for a figure like NVIDIA's R&D spend as a percentage of revenue returns the number with its history, and on the Max tier each figure is clickable back to the exact page of the SEC filing where it originates. I traced several figures through that audit path and the link landed on the right source paragraph rather than a generic filing URL. That design turns the copilot from an oracle into a research assistant you can check — and in a domain where a hallucinated revenue number is worse than no number, it is the single most important feature in the product. On lower tiers the same answers arrive without the deep-link provenance, which is a rational upsell but means the safety property is paywalled.

Second observation: the free tier is a genuine test drive but clearly fenced. Ten years of annuals is enough for a trend check, and the one-earnings-event allowance gives you a feel for transcript answers. But the 30-row dashboard cap and two years of segment data run out quickly if you follow more than a couple of names, and there is no export anywhere below Max. In my use, the free plan is best spent checking whether the Copilot answers questions in your research style at all — a week is enough to know.

Third: the MCP server is the most interesting surface for power users. Wiring Fiscal.ai into Claude as a data source means questions like comparing inventory turnover against an industry average resolve against structured data instead of a language model's memory. Configuration is documented for several clients, and the OAuth flow is standard. The caveat is API history depth: transcript and news history in the API is months, not years — news coverage starts in late April 2026 and transcripts hold roughly the latest period onward — so anything backtesting or historically exhaustive needs the terminal or another source.

A word on who loses to this product. Before tools like this, a retail investor's honest options were a $27,000-a-year terminal, a broker's shallow research notes, or manually scrolling EDGAR. Fiscal.ai compresses that middle ground: not the full terminal, but enough structured history and provenance to do real independent work. The equity research analysts I trust most use a terminal at work and something like this on personal accounts, and the split makes sense — you pay the terminal tax for execution, real-time data and compliance features, none of which Fiscal.ai attempts. Knowing that boundary is the difference between being satisfied with the product and feeling cheated by it.

What it does not do

There are no real-time quotes anywhere in the product; market data is delayed, and there are no options or futures. This is a fundamentals and filings product, so if your workflow is technical trading, level 2 data or derivatives, Fiscal.ai is irrelevant to you. Financial-data export is Max-only and covers the US, Canada, ADRs, the UK and the EU — a smaller footprint than the 100,000-plus stock coverage headline suggests. The AI Copilot answers questions; it does not execute trades, place orders, or give personalized investment advice, and nothing in it replaces a licensed advisor. The company itself is young — launched as FinChat in 2023, Series A funded in mid-2025 — so data operations maturity and uptime track record are thinner than at a Bloomberg or Refinitiv. And the unpriced, annually-billed API product means the cheapest honest infrastructure cost is unknowable without a sales call.

Smaller friction points from use: delayed data is labeled but the labelling could be louder mid-session; the segment and KPI history limits differ by tier in a way that takes a minute to internalize; and because coverage exceeds the export footprint, you can research a name you cannot extract.

Verdict

Fiscal.ai is the finance chatbot done right — a real data platform with an AI layer and provenance on tap — and for fundamental research on covered markets it competes far above its price point. I rate it 3.9: core functionality depth is strong with the audit-to-filing flow as a genuine differentiator (40%), value is good against terminals that cost 20x more (20%), usability is clean across web and MCP surfaces though tier fences are felt quickly (20%), support and ecosystem are growing but young (10%), and reliability signals are decent for a Series A startup with transparent data sourcing but little operating history (10%). Retail investors who want answers they can verify will like it; professionals who need export and auditability should budget for Max.

Sources

• https://fiscal.ai
• https://stockmarketstack.com/tools/fiscal-ai
• https://discover.oreateai.com/discover/the-true-cost-of-fiscalai-2026-pricing-and-feature-breakdown
• https://ryanoconnellfinance.com/fiscal-ai-review
• https://ryanoconnellfinance.com/partner_discounts/fiscal-ai-promo-code
• https://yespress.io/products/finchat
• https://aitoolfit.com/compare/tradingview-vs-fiscal-ai.html
• https://www.fiscal.ai/pricing