Data pipeline
MCP direct to Financial Modeling Prep (Ultimate tier). No search-engine scraping, no PDF roulette — every number is fetched from the source the sell-side uses.
For every company you follow it keeps a living thesis and a conviction score that updates as new evidence lands — and every call is logged against what actually happened. Not a chatbot with web search. A research analyst that remembers.
/d dashboards · /d deep dive · /e earnings · /s schedule · /u wiki — and more
Each company in our base is wired to its peers, supply chain, sectors, themes, and the narratives the market is pricing in. The agent doesn't reason about NVDA in isolation — it sees the AI capex cycle, the HBM supply chain, the agentic-AI supercycle narrative, and every other company they touch.
That's the difference between a chatbot with web search and a research analyst with a knowledge base.
An investment-bank analyst's thinking, encoded. Each report averages ~10 million tokens of research and reasoning before the first word is written — synthesizing historical fundamentals, financial data, the company's operating history, and the live market narrative into a single document. Not a long article; a comprehensive research report.
Type what you want. The agent queues a background build, gathers the data, lays out the panels, and notifies you when it's ready. Save the layout, share with your team, or keep iterating.
Type any question. The system spins up a background task, queues your build, and notifies you when it's ready. Walk away — come back when it's done.
The gap doesn't come from a stronger model — vanilla Claude Opus 4.7 and Agentic Capital run on the same base. It comes from the harness: how data gets in, what the agent knows about each company, and what frameworks tell it when to stop reasoning and start verifying.
MCP direct to Financial Modeling Prep (Ultimate tier). No search-engine scraping, no PDF roulette — every number is fetched from the source the sell-side uses.
Every company carries a deep dive, thesis, and wiki. News and quarterly results write themselves in automatically; the agent reads them before it answers.
Pre-deployed analytical frameworks tell the agent when to stop reasoning and go verify. The harness enforces the discipline a junior analyst takes years to learn.
Same Claude. Just taught how to be a diligent financial analyst.
| Task | Vanilla | AC | Δ (pp) |
|---|---|---|---|
| Number retrieval | 41.7% | 100% | +58.3 |
| Financial modeling | 40% | 86.8% | +46.8 |
| Trend analysis | 36% | 69% | +33.0 |
| Qualitative retrieval | 47.2% | 79.6% | +32.4 |
| Numerical reasoning | 33.3% | 62.5% | +29.2 |
| Adjustment analysis | 46.5% | 71.5% | +25.0 |
| Beat or miss judgment | 55.1% | 71.4% | +16.3 |
Vals AI Finance Agent v1.1 · 537 questions · both runs on Claude Opus 4.7 · Vanilla configured with WebSearch + WebFetch per Vals reference.
The system runs in the background — deep dives, earnings reviews, new-idea screens — and drops the results in your inbox. Plus custom alert rules across whatever dimensions you care about: Fear & Greed thresholds, QQQ RSI levels, your holdings, the macro tape.
Numbers come from FMP — the same institutional-grade financial database the sell-side licenses. Every claim is reviewed by our backend AI investment committee before it ships. Hover any figure to see source, timestamp, and revision history.
Every memo you read, every position you size up, every alert you mute — the system gets quieter and more precise. Your interests, your radar, your blind spots, learned.
Telegram, Discord, Slack, email, webhook, in-app — the agent delivers ideas, alerts, and finished reports to wherever you actually pay attention. Your team channel, your phone, a bot of your choosing.
Start with an AI chat that already knows the market. Upgrade to Power for proactive features — daily briefs, deep dives, custom dashboards, and alerts that work while you sleep.
The most comprehensive AI investment chat.
14-day free trial · card required · cancel anytime
Try free →Proactive research that works while you sleep.
14-day free trial · card required · cancel anytime
Try free →For the professional desk.
14-day free trial · card required · cancel anytime
Try free →14-day free trial on every plan · card required, cancel anytime. Team seats on Max. All tiers include the full knowledge graph and the entire dashboard library.