Mercata

How it works

Your fund's knowledge, put to work.

Mercata brings your fund's internal data and its whole investable universe into one record, inside your fund, and puts AI models and workflows to work on it.

The architecture

Any model can summarize the market. Mercata reads it the way your fund does.

Mercata has three layers, and all of them run inside your fund: the buy-side workflows the team uses, the harness that puts a model to work, and your fund's knowledge underneath. Data flows in from your fund's own systems and its investable universe. The AI model is called from outside and can be swapped at any time.

What Mercata brings in

Your fund's internal data

  • Email, Teams and Slack
  • Meeting transcripts
  • Memos and notes
  • Portfolio positions

Your fund's universe

  • Filings and transcripts
  • News and market data
  • Sell-side research
  • Alt data
  • + anything else your fund wants

Years of history, updated in real time

MercataRuns inside your fund

Work done in Mercata

Buy-side workflows

  • Thesis tracking
  • Event triage
  • New idea generation
  • Agents for any task
Every use writes back

The harness

Puts the model to work

  • Picks exactly what the model sees
  • Sets what it can touch
  • Watches the book around the clock
  • Writes back what was learned

Your fund's knowledge

Every name, thesis, position and decision, with the why

  • Every answer cited to its source

The AI model

Any model: Claude, GPT, Gemini and others

  • Called for each task
  • Swap it when a better one ships
  • Never trains on your fund's data
Mercata is your fund's own knowledge, and the layer that protects it and puts AI to work on it.

The layers

What each part does.

01 — What comes in

Your fund's own data, and its whole universe.

Mercata ingests two kinds of data, with years of history, and keeps both current in real time.

  • Internal data. Email, Teams and Slack, meeting transcripts, memos and notes, and portfolio positions.
  • Your fund's universe. Filings and transcripts, news and market data, sell-side research and alt data for every name your fund could own, not only the ones it holds.
  • Anything else. If a system has an API, an MCP server, an FTP feed, email delivery or OAuth sign-in, Mercata can bring it in.

Capture happens in the background. The team keeps working in the tools it already uses.

02 — Your fund's knowledge

Everything your fund knows, turned into a data layer AI can use.

Mercata turns your fund's internal and external data into one structured layer: what your fund believed, why and when, linked to the filings, transcripts and news on every name in its universe. A filing is linked to the names it affects. A meeting is linked to the thesis it changed.

  • The market, through your fund's lens. News and filings are read against your fund's own positions and theses, so what matters to you comes first.
  • Built for AI to use. Because the data is already structured and linked, the harness hands the model a precise slice instead of hundreds of pages. Answers come back faster, more accurate and at lower cost than having a model trawl through raw documents.
  • Point-in-time history. What your fund knew, what it decided, and when.
  • Cited. Every answer traces back to its source document.

03 — The harness

Puts the model to work.

The harness sits between your fund's knowledge and the AI model. It turns a general-purpose model into one that works for your fund.

  • Picks what the model sees. For each task, it selects exactly the right part of your fund's record.
  • Sets what it can touch. Access follows the controls your fund decides.
  • Watches the book around the clock. It flags the events, filings and news that matter to your fund's positions, without waiting to be asked.
  • Writes back what was learned. The record grows as the team works.

04 — Buy-side workflows

Where the team works.

The team does its research in Mercata, and every use writes back to your fund's record.

  • Thesis tracking. Why your fund owns each name, and what has changed since.
  • Event triage. Overnight filings, transcripts and news, sorted by what matters to the book.
  • New idea generation. Every new name starts from what your fund already knows.
  • Agents that work the way your fund would. Hand off any research task and the agent does it with your fund's own knowledge and standards: a dossier on a company you have never covered, a red-flag check before a new position, a brief ahead of earnings.
  • Ask Mercata. Talk to your data. Ask anything about your fund's names, theses and history, and get answers grounded in what your fund knows, with citations.

05 — The AI model

Any model: Claude, GPT, Gemini and others.

Mercata calls a leading model for each task. You can switch models when a better one arrives.

  • Works with models from Anthropic, OpenAI, Google and others.
  • Your fund's data never trains a model.
  • Better models improve your fund's work the day they arrive.

Clear lines

What Mercata is, and isn't.

Mercata is

  • Your fund's own knowledge, in one cited record.
  • The layer that puts AI to work on it, inside your fund.
  • A tireless colleague for every analyst: in every meeting, forgets nothing.

Mercata isn't

  • An AI model. It uses the best available and lets your fund switch.
  • A data feed. It reads data through what your fund believes.
  • A replacement for analysts. It captures knowledge and offers judgment when asked. Your fund makes the call.

Questions

Common questions.

How is Mercata different from using ChatGPT or Claude directly?
Used directly, a chat app isolates each person's knowledge in their own chats and ties it to one AI model. Mercata keeps one shared record across your fund and puts any model to work on it, so what the fund knows is useful immediately, can be built on, and stays protected. When an analyst leaves, what they knew stays with your fund.
Which AI models does Mercata use?
Leading models from Anthropic, OpenAI, Google and others. Your fund can switch models at any time without losing its knowledge.
What data does Mercata ingest?
Your fund's internal data (email, Teams and Slack, meeting transcripts, memos and notes, and portfolio positions) and its investable universe (filings and transcripts, news and market data, sell-side research and alt data), with years of history, updated in real time.
What can Mercata's agents do?
Any research task your fund defines, done the way it would do it, because the agents work from its own knowledge. Examples include a dossier on a company you have never covered, a red-flag check before a new position and a brief ahead of earnings. The agents do the work. Your fund makes the call.
What systems does Mercata connect to?
Any system your fund can reach through an API, an MCP server, an FTP feed, email delivery or OAuth sign-in, including email, Teams and Slack, meeting transcripts, document stores and portfolio positions.
Does Mercata train AI models on our data?
No. Your fund's data never trains a model and is isolated from every other firm's data.
Do analysts have to change how they work?
No. Mercata captures from email, chat, meeting transcripts and documents in the background. Work done in Mercata's own workflows writes back directly.
Does Mercata make investment decisions?
No. Mercata captures your fund's knowledge, flags what matters and offers judgment when asked. Your fund makes every call.
Who is Mercata for?
Any fundamental investor. Most typically, fundamental long/short and multi-strategy hedge funds.

See it in action

Everything your fund knows, working only for you.

15 minutes to see it on a synthetic long/short book.

Book a 15-minute demoHow we protect your data