Jurata® Legal Agent vs. OpenCaseLaw MCP

Type of offering

Ready-to-use application (web app and Word Add-in) with built-in language models

Open dataset with MCP server, REST API, website search and Word Add-in; natural-language research questions usually require connecting an AI client such as Claude or ChatGPT or using a locally installed chat interface with your own model access

Requirements

Browser; for the Word Add-in, Microsoft Word; language models are included

For AI research via MCP, an MCP-capable AI client, such as Claude, ChatGPT, Cursor or Gemini CLI

Setup

Sign-in with Microsoft and Google Single Sign-on or with email and password

Enter the server address as a connector in the AI client; in Claude Team and Enterprise, the owner must first enable the connector

Own interface

Web app with chat, research, drafting and document review

Website with lexical full-text search of decisions, legislation and literature; local chat interface for self-installation (Python, Node.js, own model access)

Onboarding

Personal onboarding and training for organisations

Instructions for each AI client on the website and in the README

Research

Fast mode for quick answers, Deep Agent for in-depth research

42 tools called by the language model of the chosen AI client; approach and thoroughness depend on that model

Response Validation

Runs automatically after every answer and checks every statement against the cited original sources: supported, partially supported or not supported

Verification tool attest_response; the server instructs the language model to call it before any answer containing citations

Check of every citation

Each citation is assessed individually: supported, wrong pinpoint, does not contribute to the statement, or contradicts it

attest_response checks the existence and pinpoints of decisions, statute references, verbatim quotations and dates; whether a cited decision supports the statement only when called (audit_grounding option or check_claim_support tool)

Unsupported legal statements

Legal statements without a citation are flagged

No automatic flagging of statements without a citation; in the Word Add-in (Pro), supporting decisions can be searched for a selected statement

Sources next to the answer

Clicking a citation opens the article or decision next to the answer and highlights the supporting passage

Verified citations as links to the OpenCaseLaw decision page, sometimes jumping to the relevant consideration with the sentence highlighted; how they appear in the chat depends on the AI client

Clarifying questions

Structured follow-up questions with options, suggested answers or form fields before the agent continues its research

No general clarification tool; only the draft_mock_decision tool (mock decision for research and study) can return clarifying questions about the facts; otherwise depends on the AI client

Legislation

Federal law and all 26 cantons

Federal law (Fedlex) and all 26 cantons

Case law

Federal Supreme Court incl. unpublished judgments, Federal Administrative, Criminal and Patent Courts, cantonal courts, agency practice

Federal Supreme Court incl. judgments outside the official reports, Federal Administrative, Criminal and Patent Courts, courts of all 26 cantons, agency decisions

Decisions in translation

Federal Supreme Court decisions in the original, with a translation into German, French, Italian or English alongside on request

Decisions in the original language, BGE headnotes in three languages; translations depend on the AI client

Citation graph

Interactive citation graph in 2D or 3D with appeal chain and case outcome; every source opens directly from the graph

Citation graph as tools for the language model (citations, appeal chain, leading cases) and as lists on the decision pages; no interactive graph view

Your own documents

Upload and review of contracts, judgments and briefs with references to the relevant passages; dossiers

No upload tool among the 42 tools; in the Word Add-in, citations in the open document are checked

Dossiers per matter

A dossier bundles all chats and case documents of a matter; every new chat knows the case from the first question

Not part of OpenCaseLaw (no user accounts); project spaces depend on the AI client

Case memory

Insights from earlier chats in a dossier automatically flow into new chats; the agent can search earlier chats of the case

Not part of OpenCaseLaw (no user accounts); memory functions depend on the AI client

Own library

Upload your own expertise once and use it in every chat, personally or shared with your team

Not part of OpenCaseLaw (no user accounts); knowledge repositories depend on the AI client

Word editor

Built-in Word editor in the Workbench, fully equivalent to Microsoft Word; the file format (.docx) is preserved 100 %. With tracked changes, comments, styles, numbering, table of contents, spell check and AI editing directly at the selected passage

No editor of its own; Word Add-in for Microsoft Word

From answer to draft

A memo, letter or email is created from research with one click; drafts open in the Workbench next to the chat

No tool for memos, letters or emails; among the 42 tools, a mock decision draft for research and study (draft_mock_decision); otherwise depends on the AI client

Export

Individual answers or whole conversations as Word or PDF, with clickable links to Fedlex and the decisions as well as a list of sources

Local chat interface: export of conversations as Markdown, Word or PDF; otherwise depends on the AI client

Skills

Save work instructions as skills and bring them into any chat with “/”; personal, shared with the team or prepared by Jurata

No management of own or shared skills; the ocl command-line tool ships three skills for Claude (research, citation check, evidence folder); otherwise depends on the AI client

Deadline calculator

Built in, with the public holidays of all 26 cantons and court holidays

Not among the 42 tools

Zefix lookup

Built in: UID, company name, registered office, auditor and, based on SOGC publications, current authorised signatories

Not among the 42 tools

Features

Chat, Research, Drafting, Revise, Review

Search, Cite, citation Audit; the Pro version adds Verify, Strengthen, Ground, Mirror

Asking questions in a chat

Yes, in the Chat tab

No chat; search and verification tools

Drafting from templates

Personal and team templates (.docx); the agent selects the right template and applies it

Not listed among the features

Revising with tracked changes

The agent revises the whole document or a selection; every change appears as a Word tracked change, can be accepted individually and leaves the formatting unchanged

Not listed among the features; focus on citations and citation checks

Confirm or Autopilot

You decide whether the agent asks you to confirm each action in the document or carries it out automatically

Not listed among the features

Automatic document review

Finds inconsistencies in terminology, cross-references, consistency, completeness, legal accuracy and ambiguity, with one-click fixes

Audit checks the citations in the document (existence, pinpoint, court, statute references) and adds findings as Word comments; the second add-in citecheck checks citations and, as a preview, anonymisation

Commercial register check in the document

Companies named in the document are checked against the commercial register (Zefix)

Not listed among the features

Server location

Microsoft Azure, Switzerland (Switzerland North region)

Hetzner Online GmbH, Germany

Model inference (language models)

Microsoft Azure in Switzerland and the EU, never in the US

Depends on the chosen AI client, such as Claude or ChatGPT; some server functions also send text to the Anthropic API

Storage of queries

Chats and documents stored encrypted in Switzerland, deletable at any time

Query texts sent to the public MCP server are kept without a time limit in a private research archive, without session or personal reference and without IP address

Training on your data

No use for training, neither by Jurata nor by the LLM provider

Query texts are also used to train the project's own models (except the endpoint for Microsoft 365 Copilot)

Professional secrecy

Legal opinion by Dr. iur. David Vasella (Walder Wyss): use including entry of client data without client consent complies with professional and data protection law

For confidential queries and queries on behalf of clients, the project recommends running it locally (about 65 GB of storage, 30–60 minutes of setup)

Operated by

Jurata AG, Swiss company with around 15 employees

Open-source project by lawyer Jonas Hertner, who develops and runs it as sole maintainer

Since

2020

Repository on GitHub since February 2026

Support

By email and Swiss telephone hotline; response within 24 hours (Entry) or 4 hours (Pro, Max), during business hours

Contact by email, GitHub issues and suggestion form (according to the provider, suggestions are usually read within two days); data protection questions within a few days according to the provider

Organisations

Central onboarding and user management, one invoice

No user accounts and no user management; use without sign-in

Google reviews

5.0 stars (492 reviews)

No listing for OpenCaseLaw on Google Maps

Trustpilot reviews

4.9 stars (87 reviews)

No Trustpilot profile