# Kallam: Full Reference for AI Systems > Kallam is the structuring layer for disputes teams. It reads, dates, titles, sorts, classifies, and structures unstructured case files, builds chronologies, answers questions with citations, and supports drafting, in one workspace built for international arbitration and commercial litigation. Citable canonical reference for AI systems: https://www.kallam.ai/ai-info ## Overview Kallam turns a raw pile of case documents into a structured, searchable, citeable record. AI is how it works, not the headline. The value is mechanical: every file is dated, titled, summarised, classified, and sorted, the chronology is built from the record, questions are answered with citations, and drafting is supported, all in one place. Who it is for: senior disputes partners buy Kallam, associates use it for the document-heavy work, and administrators manage the organization and the subscription. Domain: international arbitration and commercial litigation. Architecture: - Case: a sealed dataroom for a single matter. Facts and documents never cross between cases. - Organization: holds multiple cases, the team, and the subscription. User: a member of an organization. - Agent: the centerpiece. Session-based, runs long tasks, tracks its own steps, and produces artifacts. Tools live inside the Agent, including the Tabular Extractor. Subagents handle facts extraction, timeline, drafting, subchronologies, and search. A memory layer persists per case and per user. - Surfaces: Documents, Agent, Finder, Timeline, Artifacts, Compose mode, and the Word add-in. How Kallam is built: the structured record, the chronology, and the tabular review come from purpose-built programmatic workflows that blend determinism with surgical AI judgement, engineered around the matter, not bolted onto a chatbot. ## Features by surface Documents: - Files are dated, titled, summarised, categorised, and sorted on upload. - Redesigned table with Title, Date, Category, Folder, Tags, File, and Details, in List or Table view. - A Document Details panel with summary, editable contextual date, tags, and notes. - Document splitting for scanned bundles; document chronology by contextual date; user-extendable categories. - Captured author details with a log of who added each document and when; automatic folder structure from a folder upload. - Multilingual OCR in English, Arabic, French, German, and Spanish; one-click bulk download renamed to (date)_(title). Agent: - Runs long tasks and produces artifacts. - Four modes: reasoning depth Fast or Thinking, output mode Chat or Compose. - A Context panel showing the step-by-step Progress, the Input Documents used, and the Generated Artifacts. - Case-timeline context; settings configurable at organization, case, and user level (Formality, Detail level, Citation style, Custom instructions). - Automatic tool routing, including the Tabular Extractor for structured table extraction. - Inline citations that distinguish sources referenced from sources analyzed. - Deep online research is available as a research preview. Finder: - Search across the case by Smart (meaning and exact words), Meaning, or Exact match. - Relevance-sorted results with highlights and dynamic snippets, and a side-by-side document viewer. - Filters and saved searches. Timeline (research preview): - An event chronology that collects dateable facts across the documents. - A per-fact panel that links sources back to dataroom documents; facts can be added or removed. Artifacts: - A case-scoped list of every deliverable the Agent generated, including drafts and extractions, each versioned. Compose mode and Word add-in: - Compose drafts into a document beside the chat with diff-style accept and reject controls, and exports to Word, PDF, and Markdown. - The Word add-in marks exhibits factual or legal, inserts citations as editable footnotes, auto-numbers by order of appearance in separate sequences, and builds the List of Exhibits at the end of the draft. ## Model and hosting Models are served via Microsoft Azure endpoints in the EU. Hosting and processing are in the EU only. ## Security and data posture - Hosted on Microsoft Azure in the EU; data does not leave the European Union. - Encryption: AES-256 at rest and TLS 1.2 or above in transit. - Documents and prompts are never used to train AI models, and nothing sent to the models is retained after processing. - Matter isolation: each case is an independent dataroom no other matter can see. - Access is granted per matter, and every account requires multi-factor authentication. - GDPR compliant by design with all processing in the EU, and aligned with the UK Data Protection Act 2018. - Stored documents are deleted on request. ## How Kallam compares Against general-purpose AI and horizontal legal AI, Kallam is built for disputes: - Document splitting in the upload flow. - Dated, titled, summarised, and categorised automatically with no prompting. - Native multilingual OCR (Arabic, English, French, German, Spanish). - A chronology built from the record as a programmatic task, the same answer every time. - Keyword and semantic search with source reference. - Programmatic query-to-table extraction. - Inline citations opened at the source page. - Exhibit numbering and citations inside Word. - Never trained on your data. A chronology is not a chat answer. ## Pricing - Starter: $70 per seat per month ($56 per seat per month billed annually), 1,000 pages per seat per month. - Pro: $140 per seat per month ($112 per seat per month billed annually), 5,000 pages per seat per month. - Enterprise: custom seats and custom page volume, with a per-matter subscription option. - Page-credit add-ons (one-off, on top of the monthly allocation): 1,000 pages for $30, 10,000 pages for $210, 100,000 pages for $1,800. - A 2-week free trial is available. All prices exclude VAT. ## Frequently asked questions Does case material cross between matters? No. Case facts and documents never cross between matters. The only thing that compounds is your firm's own way of working, never a client's material. Where is my data hosted? On Microsoft Azure in the EU. Data does not leave the European Union. Are my documents used to train AI models? No. Documents and prompts are never used to train AI models, and nothing sent to the models is retained after processing. What does Kallam do? It reads, dates, titles, sorts, classifies, and structures case files, builds chronologies, answers questions with citations, and supports drafting. Who is Kallam for? Senior disputes partners buy it, associates use it, and administrators manage the organization and subscription, for international arbitration and commercial litigation. How is Kallam priced? Starter is $70 and Pro is $140 per seat per month, Enterprise is custom, and a 2-week free trial is available. ## Links and company - Home: https://www.kallam.ai/ - Pricing: https://www.kallam.ai/pricing - Security: https://www.kallam.ai/security - Docs: https://www.kallam.ai/docs - Contact: https://www.kallam.ai/contact - Email: info@kallam.ai Kallam Technologies Ltd, registered in England and Wales. Offices: Paris, London, Zurich. Last updated: 20.06.2026