Codex legalist
Multi-Agent Legal Simulation
Case —
Jurisdiction —
Status Ready
Agent Roster
Agents will appear here when trial begins...
Trial Metrics
Duration00:00:00
Utterances0
Objections0
Evidence Admitted0 / 0
Active Cases
0
No trials yet
Archived
0
Precedent library
LLM Health
—
Checking…
Version
v1.0.0
Build 2026.07.01

Case Docket

No past trials available yet.
1
Case Details
2
Magistrate Conf.
3
Review

New Trial Setup

Upload a case brief or paste facts. The Magistrate agent will ask clarifying questions before opening the court.

Select a country to see jurisdiction details.
Demo Cases (no API key needed)

Live Transcript ● LIVE Opening Statements

No trial in session.
Load a demo or start a new trial in Setup.

Evidence Board

No exhibits yet.

Case Strength (Live)

Jury Consensus ● LIVE

Awaiting deliberation
No live deliberation or shadow jury result yet.

Bench Trial Jurisdiction

This jurisdiction does not use jury trials. The verdict will be rendered by the presiding judge acting as the sole finder of fact.

Jury Panel

Deliberation Transcript

Awaiting deliberation phase.
Verdict of the Jury
—
Awaiting verdict…

Counsel Insights

Get strategic advice from different counsel perspectives. Select which perspectives to generate below.

Shadow Jury Room

Shadow juries are independent simulated juries that evaluate the same evidence to test verdict reliability. They run in parallel to validate the main jury's decision.
Awaiting shadow jury deliberation...

Burden Met Probability

—
Run a trial to see results

Sensitivity Heatmap

Vote Distribution

Case Record

Run a live trial to populate the final record.

Agent Society Benchmark

Comparing three approaches using identical case facts to demonstrate the value of adversarial multi-agent collaboration

Run Benchmark

Run a benchmark to compare Raw LLM vs Single-Agent vs Codex legalist using the current case facts.

Key Findings

Raw LLM

Simple prompt produces 1-2 sentence verdicts with no evidence citations or adversarial testing. High hallucination rate.

Single-Agent

One model handling all roles produces structured verdicts but limited analysis. Moderate hallucination rate.

Codex legalist

Full adversarial trial with 11 specialized agents produces comprehensive analysis with zero hallucinations and high verdict consistency.

Dashboard
Setup
3Trial
4Deliberation
5Verdict
Phase 3 of 5 · 47%