Pascal Data Agent

Turn your firm's files into data your analysts and agents can use.

Pascal Data Agent reads broker reports, expert calls, internal notes and email, gives each file a type, extracts its fields and tags it by company. Analysts browse the result in the vault viewer. Agents query it over MCP.

99.8%

Automation rate

50M+

Pages indexed

35K+

Stocks & entities mapped

How it works

From file to fact in four steps.

01Give every file a typeEach file gets a type from your taxonomy, with a confidence score. Unsure files wait for a person.
Mgmt_mtg_0930.pdf
Broker report0.97
Expert call
02Pull out the fieldsRating, price target, speaker, relationship, stance. Each value keeps the page it came from.
Broker report fields
RatingBuyp.9
Price target$245p.9
03Resolve to the context graphTickers, name variants and product lines map to one company in your context graph.
Entity resolution
Blackwell Ultra→NVDA
Nvidia→NVDA
04Serve people and agentsAnalysts open the vault viewer. Agents call the same records over MCP, under the same permissions.
Served to
Vault viewerAnalyst UI
list_documentsMCP

Vault viewer · Analyst UI and Agent (MCP)

One dataset, two ways in. Try it on a sample vault.

Pick a document type, add a company, change the date range and click a row to open the file at the section about that company. Switch to Agent (MCP) to see the same records as list_documents and get_document return them. Sample data is illustrative.

See MCP & APIs

Document classifier

Every file, sorted by type. Each correction teaches it.

Each file gets a type from your firm's taxonomy, with a confidence score and a runner-up. Files below your threshold wait in a review queue. One click assigns a type, Pascal re-scores similar files, and each type starts its own workflow.

See the workflow library

Extraction & table editor

Every field and table, traced to its page.

Each document type has a schema: rating and price target for broker reports, covenants for credit agreements, speaker and relationship for expert calls. Tables come out of PDFs as tables. When an analyst corrects a cell, Pascal keeps the original, the editor and the page region.

See how agents read each value

Vertical entities

The entities a general model walks past.

Speaker names on expert calls. Brokers. Ticker variants and product lines that belong to a company. Ratings and price targets. Pascal tags them the way an analyst reads them and resolves each one to a single company in your context graph.

See the context graph & memory

Permissions

Two analysts, two different vaults.

Lists, counts, answers and MCP calls follow each user's licences and source permissions, including per-broker research entitlements. The viewer says how many files it held back and why. Pascal runs in your own cloud or on premise.

See how Pascal handles security

Is Pascal a data company?

A data platform for your own research.

Pascal is a data platform for your own research, the way Snowflake is a data platform for your tables. Your files stay yours. Pascal gives them a schema, keeps them in sync and serves them to the tools you approve.

See the Pascal platform

Built for the files your firm already has.

Frequently asked questions

The part of Pascal Agents that turns your firm's files into structured data. It classifies each file, extracts the fields and tables that matter for its type, links them to your context graph and serves them to analysts and agents.

See your own files
as structured data.

USA

228 Park Ave S PMB 55152, New York

UK

86-90 Paul Street, EC2A 4NE, London

India

Indiqube Edge, ORR, Bangalore

Copyright © 2026 | All Rights Reserved

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.