Semplifica. — product suite

MikeRust logo
MikeRust

Open-source legal AI. Runs on your machine.
No cloud dependency. No lock-in. Built for Europe.

Document review Tabular review Projects vault Claude · Gemini · Local LLMs EUR-Lex · Italian law corpus AGPL-3.0 · Free
The application

A desktop app that runs
entirely on your machine.

MikeRust is a Tauri desktop application: a Rust+axum backend and a Next.js frontend running in the same process, on your own computer. No server to provision, no cloud database, no external auth provider. Everything lives in a local SQLite file and a local storage folder.

Outbound traffic happens only when you explicitly call a remote LLM (Anthropic, Gemini, OpenAI) or a remote MCP server you configured yourself. No telemetry, no anonymous metrics, no remote logging — ever.

Chat with document citations and inline PDF viewer
Tabular review — bulk contract analysis in a structured grid
Projects vault — files, colleagues, persistent context
Local RAG — sync folders, embed with ONNX, search offline
EUR-Lex and Italian legal corpus built in
MikeRust · desktop app · local instance
●  Document assistant
●  Tabular review
●  Projects vault
●  EUR-Lex · Italian corpus
axum :3001 · Next.js :3000 · mike.db · local
The problem

Harvey costs a fortune.
MikeOSS is a great start.
Neither is built with Europe in mind.

Harvey and Legora are valued at billions — but they process documents on US-controlled infrastructure, operate under US law, and charge enterprise contracts that price out most European firms. MikeOSS proved that their core functionality can be rebuilt openly. MikeRust takes that proof and makes it production-ready for the European market.

The GDPR requires data to stay under the controller’s authority. The EU AI Act imposes new obligations on high-risk AI in legal contexts. US-based SaaS providers cannot structurally satisfy these requirements — regardless of what their DPAs say.

MikeRust solves this by design: a desktop app that runs entirely on your machine, connects only to the LLM provider you choose, and keeps every document, query and embedding under your control — in a single local SQLite database.

Harvey · Legora — $11B & $5.5B valuations
VS
Enterprise contracts pricing out small & mid firms
US-hosted infrastructure, US law applies
VS
GDPR data residency & sovereignty requirements
MikeOSS — excellent, cloud-native (Supabase + S3)
VS
European firms need local-first, no external services
✓ MikeRust resolves this tension by design
How it works

From install to first review.
Four steps. No cloud.

Everything runs on your machine. No files are transmitted to external services. Your documents stay under your organisation’s control from start to finish.

1

Install the desktop app

MikeRust ships as a native Tauri desktop application for Windows, macOS and Linux. Runs on standard hardware — no GPU or NPU required, but automatically leverages DirectML (Windows GPU/DX12) or Qualcomm QNN (Snapdragon NPU) acceleration when available.

2

Configure your environment

Set your LLM provider API key and a JWT secret in a .env file — everything else has sensible defaults. On first launch, MikeRust creates data/db/mike.db and downloads the local embedding model (~280 MB, one-time).

3

Add your documents

Upload files via chat, sync local folders for RAG, or connect authoritative legal corpora — EUR-Lex (24 EU languages) and the Italian legal corpus (Normattiva + Corte Costituzionale, ~90K acts) are built in. PDF, DOCX, RTF, XLSX, TXT and CSV supported natively, with no external process spawned.

4

Your data never leaves your machine

All documents, embeddings and chat history live in a single local SQLite file. Outbound traffic only when you explicitly call a remote LLM. No telemetry, no usage reporting. Encrypted project bundles (.mikeprj, AES-256-GCM + Argon2id) for secure cross-machine sharing.

Features

Everything a legal team needs.
Nothing they don’t.

Core
Document AI

Review, draft and edit legal documents — contracts, memos, pleadings, due diligence reports. Citations with inline PDF viewer: click a source pill and jump to the exact highlighted passage.

  • Chat with cited sources
  • Inline PDF viewer
  • Tracked-change detection (DOCX)
Bulk analysis
Tabular Review

Analyse dozens of contracts in a structured grid. Extract clauses, flag deviations, compare across documents. Export to CSV or JSON for downstream use in your own systems.

  • Bulk document intake
  • Configurable columns
  • CSV / JSON export
Collaboration
Projects Vault

Organise matters into projects. Upload files, share with colleagues, keep context across sessions. Encrypted .mikeprj bundles for secure cross-machine sharing — keys derived via Argon2id.

  • Matter-based organisation
  • AES-256-GCM project export
  • Hash-keyed attachment cache
Knowledge base
Legal Corpora

Sync EUR-Lex (24 EU languages, CELEX/ELI lookup) and the Italian legal corpus (Normattiva + Corte Costituzionale, ~90K acts) directly into your local vector index. Search stays fully offline.

  • EUR-Lex — 24 EU languages
  • Normattiva + Corte Cost.
  • Offline vector search
Technical edge

Rewritten from the ground up.
Compact, fast, dependency-free.

01

Rust + axum backend

The backend is pure Rust (axum), embedded in the Tauri process alongside the Next.js frontend. No LibreOffice, no Supabase, no S3. PDF via pdfium-render, DOCX/RTF/XLSX via native Rust crates — no external processes spawned.

02

Local embeddings, hardware-aware

ONNX Runtime runs multilingual-e5-base (768 dims) entirely on-device. Execution priority: Qualcomm QNN (Snapdragon NPU) → DirectML (Windows GPU/DX12) → CPU. No GPU required; acceleration is automatic when available.

03

SQLite — one file, everything

Schema, chat history, document metadata, and vector embeddings (via sqlite-vec) all live in a single data/db/mike.db. Backup is a file copy. No database server to manage.

04

Bring your own model

Connect Claude, Gemini, OpenAI, vLLM or Ollama. Vision-capable models automatically handle scanned PDFs and images. MCP client built in — connect any HTTP/SSE MCP server, including Semplifica.Edge for on-premises pseudonymisation.

05

Multilingual UI

Interface fully wired for i18n via next-intl. English and Italian shipped today; adding a new locale requires a single JSON translation file. No hardcoded user-facing strings — every string goes through a typed namespace key.

06

Sovereign data by construction

All user data lives under the local workspace: data/db/, data/storage/, ONNX weights in the user profile. PIN + Windows Hello biometric auth. No outbound traffic except explicit LLM or MCP calls.

Who it’s for

Built for the firms
enterprise SaaS prices out.

MikeRust targets organisations that cannot justify a six-figure Harvey contract — or simply refuse to hand their client data to a US-jurisdiction vendor.

⚖️

Small & mid-size firms

Solo practitioners, boutique firms and regional practices who need serious legal AI without a six-figure contract. Full feature parity with the enterprise platforms — at infrastructure cost only. No per-seat licence, no subscription.

🏛️

In-house legal teams

Corporate legal departments processing contracts, NDAs and due diligence internally. A desktop app means documents never leave the corporate machine — satisfying both GDPR controllers and IT security policy, without a procurement process.

🔧

Legal engineers & LegalTech builders

Developers and legal technologists who want a production-grade foundation. Clean Rust codebase, REST API, documented extension points, LegalCorpusAdapter trait for adding new national corpora. Fork it, extend it, ship it.

Architecture

One process.
One file. Fully local.

MikeRust is a Tauri desktop application: the Rust backend and the Next.js frontend run in the same process on your machine. No server to provision, no container required for standard use.

All documents, embeddings and chat history live in a single local SQLite file. Outbound traffic only when you call a remote LLM. No telemetry, no remote logging — by construction, not by policy.
  • ✓ Windows 10/11 — x64 & ARM64
  • ✓ macOS — native app
  • ✓ Linux — Tauri AppImage / deb
  • ✓ No GPU required · DirectML & QNN when available
  • ✓ PIN + Windows Hello biometric auth
  • ✓ MCP client — connect Semplifica.Edge and other servers
architecture
🖥  Tauri desktop window (Next.js :3000)
HTTP + SSE
⚙️  axum backend (:3001) — same process
🗄  SQLite · sqlite-vec · fastembed ONNX
📁  data/storage · .mikeprj bundles
outbound only when user invokes
☁️  Claude · Gemini · OpenAI · Ollama · vLLM
🔗  MCP servers · Semplifica.Edge
Get started

Install. Configure. Review.
Your data never moves.

MikeRust is open source and free. Clone the repository, set your LLM API key in a .env file, and run cargo tauri dev. On first launch the local embedding model downloads automatically (~280 MB, one-time). Your first document review is minutes away — with everything staying on your machine.

AGPL-3.0 — free for internal use · No usage fees · No telemetry.
git@semplifica.ai · Cesena (FC) · EU