Dockase is the unified legal AI system. Orchestrate custom AI agents, deploy domain-adapted legal reasoning, and automate regulatory compliance in one secure workspace.
Dockase Developer API
13,000+ verified judgments, citation graph, and legal embeddings over one REST API.
Trusted by forward-thinking firms
Step inside the product. Run a research question, open a landmark precedent, watch a motion come together — every surface responds instantly, exactly as it does for our clients.
Your question
Are tax matters arbitrable under Nigerian law?
Our advanced legal reasoning engine, powered by the fine-tuned Amicus-Instruct-7B model. Domain-adapted specifically for Nigerian and Commonwealth jurisprudence, it interfaces directly with our custom vector space and citation networks to draft briefs, analyze precedents, and answer complex legal queries with senior-associate-level precision and zero hallucination.
The Vault turns your firm's files — pleadings, contracts, ledgers, rules, judgments — into a private intelligence your team can question. Not a generic chatbot guessing from the internet: an assistant that answers from your documents, and cites the exact page it read. No model is trained on your data, nothing is shared — your files ground the answers, and you can verify every one.
How it works — the walkthrough
PDFs, DOCX, Excel statements — even 2,000-page bundles and phone-camera scans straight off CamScanner. One room per matter, client, or audit.
OCR for scanned pages, legal-aware chunking, and vector embeddings index every paragraph — serverless infrastructure does the heavy lifting, so nothing is skimmed and nothing is missed.
Question the whole room in plain language. Every answer arrives with its source documents and page numbers — click the citation and the PDF opens at that page.
One click runs deep analysis across every document: diligence grids, statement ledgers, fact chronologies, evidence maps — each cell and claim pinned to its page.
Your rooms become the grounding for drafting agents, research, and review — an assistant that knows your precedents, your playbooks, your files. Yours alone, by default.
Inside the Vault
Conversational Q&A over every document in the room — follow-ups included. Answers quote passages verbatim and link each claim to its document and page, so a senior partner can verify a junior's AI answer in seconds.
A question matrix per matter type — every document against every diligence question. Cells answer with page-referenced evidence, confidence flags mark what needs human eyes, and the engine surfaces contradictions between documents and what's missing entirely.
Drop in bank or financial statements (Excel/CSV) and get a categorized transaction ledger — counterparties extracted, inflows vs outflows totaled, exportable to a clean workbook. Built for audit, tax, and forensic review.
For filings and evidence rooms: a chronological timeline of the facts, each party's position, and every material claim mapped to the documents and pages that support it. The case theory, assembled from the record — not from memory.
One click sweeps the data room across the areas a deal actually turns on — corporate structure, material contracts, litigation, regulatory, employment, IP, tax, property — with findings, severity-flagged issues, and source anchors for each.
A dedicated room for your firm's knowledge — rules, articles, templates, past work — feeding research and drafting by default. Keep it private to your firm, or flip it public and contribute to the shared legal corpus every Dockase user draws from.
Firm-isolated by design
Your rooms answer only your firm — retrieval is scoped at the query level, not by policy.
Privilege-aware
Privileged documents block share links until someone with authority opts them in.
Verifiable answers
Every claim carries a document and page citation — and every AI action lands in a hash-chained audit trail.
You own your data
Files live in your private cloud bucket. No training on your documents — ever.
Every document you file in a case or contract workspace is automatically mirrored into its Vault room — no double uploads.
Even the best legal AI tools get citations wrong 17–33% of the time (Stanford). So Dockase was built to be checked, not trusted blindly — every answer carries its own verification chain.
Answers are assembled from a 26,000+ judgment corpus and your own vault — retrieval first, generation second.
A five-state verification pass — verified / verified-external / pending / mismatch / not-found — against our corpus and national legal databases. Absent from our library never means "fake" — pending means pending, and it says so.
Nothing exports until a lawyer approves it — checklist, attestation, and a Review Certificate embedded in the final DOCX.
9f2c…e1 · a47b…0c · d3e8…9f
hash-chained · exportable
Every run is hash-chained — who ran it, which model, which sources, what the guardrails found. Exportable evidence, not a promise.
Court hierarchy-aware authority ranking and overruled-case demotion are computed nightly from our citation graph.
Dockase isn't assembled from off-the-shelf APIs. We train our own models, benchmark frontier systems against each other on Nigerian legal tasks, and engineer the infrastructure ourselves — because legal intelligence for African law didn't exist, so we built it.
amicus-embed-ng-v1
General models read legal text as generic language. Ours was trained on Nigerian and Commonwealth jurisprudence, with an 8,192-token context window — entire statutory chapters and multi-page briefs embed as a single vector, so retrieval understands doctrine, not keywords.
head-to-head legal reasoning bake-offs
Before any model touches production, it competes in elimination rounds against its rivals on real Nigerian legal reasoning — drafting, analysis, citation discipline. We don't read vendor benchmarks; we run our own, and the losers don't ship.
followed · distinguished · overruled
We parse who-cites-whom across the corpus and classify the treatment — followed, distinguished, overruled — then score authority across the network. Answers don't just find cases; they know which ones still stand.
tamper-evident · exportable · verifiable
Every retrieval, generation, and human sign-off is recorded in a cryptographic hash chain your firm can export and independently verify. When AI touches legal work, "trust us" isn't a control — proof is.
Serverless research infrastructure
Document extraction and OCR run on autoscaled compute that spins up per job and vanishes — a 2,000-page scanned bundle reads in minutes, and costs nothing while idle.
Research spins out products
Our lab work became Spectra — AI-content detection and citation verification — and Dockase Scholar for law students. The platform you use is the lab's day job.
African law, first
Nigerian judgments, statutes, rules of court and practice conventions are our corpus of record — not an afterthought layered on American law.
General-purpose models see legal text as generic language. The Amicus Embedder (amicus-embed-ng-v1) matches the exact semantic weight of legal concepts, Latin maxims, and local statutory terminology with supreme domain precision.
Generic AI models represent documents as simple textual coordinates, struggling to capture the complex interdependencies of legal authority and judicial syntax.
Purpose-built for legal research, capturing legal relationships and statutory definitions with unmatched precision.
Not all citations are created equal. The Amicus Authority Graph Engine (AAGE) maps the entire web of Nigerian case law using the Amicus Jurisprudence Graph Schema (AJGS). By analyzing the deep network of citations, it identifies and dynamically elevates landmark rulings.
Elevates highly cited supreme court judgments in retrieval results, ensuring you reference the absolute authority first.
Visualizes cross-references and extracts ratios across all full-text documents automatically to trace the line of precedents.
Nigerian Case Precedents Indexed
Citation Accuracy Verified
Semantic Retrieval Latency
Automate entire legal lifecycles, not just conversations. From Client Intake to Bit-Level Contract Review, Amicus executes complex, multi-step workflows autonomously.
Pre-built or custom pipelines for Client Onboarding, Deal Closing, and Litigation Management.
Your firm's specific instructions. Clause snippets, risk tolerances, and logic rules that guide Amicus within every workflow.
A unified timeline for every matter. Filings, tasks, evidence, and client communications synchronized in one immutable record.