A living matter record, with its sources attached.
Three layers of record and one layer of intelligence. Every value the product shows points back to the fetch it came from, and only the top layer is derived.
Where the matter is, and where it is going.
Stage, days in stage, expected stage exit, next likely events, velocity, complexity, discovery intensity, expected review volume, settlement window. Each carries a confidence and the number of comparable matters behind it. There is deliberately no field for who wins.
stage: discovery stage_confidence: 0.91 days_in_stage: 12 expected_stage_exit: 2027-02-08 next_likely_events: [order.protective, discovery.deposition_notice] expected_gb: { p25: 60, p50: 110, p75: 140 } discovery_intensity: high ediscovery_trigger_at: 2026-11-24 vendor_signal: null comparable_count: 87 as_of: 2026-09-25T14:03Z // no outcome field. ever.
Computed from the entry, with the rule that produced them.
An answer due 21 days after summons returned. A 26(f) conference relative to the scheduling conference. Response windows on discovery requests. Every deadline shows its trigger entry, its rule id, and whether the court set it or we computed it.
Judges and counsel, as the record shows them.
Median days to scheduling order. Motion-to-dismiss and summary-judgment grant rates. Continuance rates. Active matters. Every figure with its count. Profiles describe timing and process; they do not predict how anyone will rule.
Fourteen tools your assistant can call.
A Model Context Protocol server over the same API. Read-only toward the court, budgeted for anything that costs money, and every response marks court-supplied text as untrusted.