LMbox
← All agents
Law firms pilot v0.1.0

JurisRecall

The firm's living memory. Unified search across internal case archive, internal doctrinal notes / memos, and external case law (Doctrine.fr, Lexbase, Légifrance). Results ranked by relevance, mandatory citation provenance, strict per-user ACL enforcement.

Measured gain
1-2 h
par dossier (vs grep manuel)

Across 100 matters/year/lawyer × 20 lawyers: 2,000-4,000 research hours recovered, ~€500k-1M of time freed. The true differentiator vs. Doctrine.fr: the firm's internal memory, accumulated over 10-20 years.

The problem

« Remember that 2019 Supreme Court ruling on excessive non-compete? » - a firm's memory is worth a fortune, exploited at 5%.

Today

Mémoire orale du cabinet

« Tu te souviens de l'arrêt sur la non-concurrence de 2019 ? »

30 min - 2 h / recherche dossier
60 % du temps : « où on l'avait géré ? »
Mémoire exploitée à 5 % sur des cabinets de 200 dossiers actifs
  • · 200 dossiers actifs + 2 000 archivés + 5 000 notes internes
  • · Recherche dépend de qui se souvient quoi
  • · Doctrine.fr / Predictice n'ont QUE la jurisprudence externe
  • · Aucun outil n'expose la mémoire interne du cabinet
L'actif unique du cabinet, ignoré
With JurisRecall

Recherche unifiée multi-corpus

Dossiers + notes + jurisprudence externe en une requête

Quelques sec. / requête
ACL respectées par utilisateur
2 000-4 000 h récupérées / 20 avocats / an
  • RAG hybride dossiers internes + ACL strictes (junior ≠ partner)
  • Doctrine.fr / Lexbase / Légifrance / Dalloz selon abonnement
  • Pertinence ★★★★★ → ★ avec citation provenance obligatoire
  • Aucun arrêt inventé, aucune référence dossier hallucinée
Pas d'invention · grounding strict
How it works

Four steps, from incoming document to human decision.

  1. 1

    Lawyer query

    A precise legal question: « precedents on a 5-year excessive non-compete clause in a tech SME ». Query logged in the audit chain.

  2. 2

    Parallel multi-corpus search

    Agent calls search_dossiers_internes (hybrid RAG + ACL), search_notes_internes (firm doctrinal memos), search_jurisprudence_externe (Doctrine.fr / Lexbase / Légifrance / Dalloz per subscription).

  3. 3

    Strict permission-checking + honest scoring

    check_acces for every internal matter before surfacing. Relevance ★★★★★ → ★ based on clause similarity, jurisdiction, counterparty. No score inflation: 3 ★★★★★ results > 20 ★★ results.

  4. 4

    Structured list + related searches

    Separate sections Internal precedents / External case law / Doctrinal notes. Mandatory provenance (matter ref + lawyer, jurisdiction + ruling number). Suggested alternative reformulations.

Architecture

Tools, connectors, deployment.

Tools (function-calling)

4
  • search_dossiers_internes
  • search_notes_internes
  • search_jurisprudence_externe
  • check_acces

Optional connectors

4
  • sharepoint
  • doctrine
  • lexbase
  • legifrance

Each connector activates based on the customer's subscription.

Deployment via the SDK
$ lmbox agent deploy ./jurisrecall \
    --box BOX-XXX \
    --token "$LMBOX_BOX_API_KEY" \
    --api https://api.lmbox.ai

LMbox guarantees across the catalogue

Data stays with you

Model and data stay on the customer's LMbox appliance. No patient, contract or invoice data is ever sent to an external cloud.

Audit chain

Every tool call, every agent output is timestamped, hashed and admissible before the regulator (ACPR, ANSM, CNIL, EBA).

Human decision

The agent recommends, the human decides. No auto-signature, no auto-payment: final responsibility stays with the business.

Try JurisRecall on the public demo.

One-click sign-in. You see the agent installed on a real LMbox, with its system prompt loaded and audit chain live.