v0.1.0 · In Development·Autonomous Systems·Primary: Rust & TypeScript

Kyro

Local-first autonomous desktop AI agent built with Tauri v2, Rust & React 19.

A high-performance, local-first autonomous desktop AI agent that bridges natural language reasoning with real computer interaction, zero-trust local privacy, and Model Context Protocol (MCP) integrations.

Official Repository Official Website Documentation Releases & Binaries

Problem Statement

Cloud-wrapped AI assistants upload your sensitive screen captures, clipboard contents, and documents to third-party servers, introducing latency and privacy risks while failing to execute real system workflows.

Intended Audience

Engineers, researchers, and privacy-conscious users seeking autonomous desktop automation without third-party data leakage.

Core Architecture & Mental Model

Multi-threaded Rust native core with Tauri v2 and React 19. All conversations, vector embeddings, task histories, and artifacts stay strictly on your local disk in SQLite (WAL mode). API keys and credentials are encrypted using OS Keyring vaults.

Verified System Capabilities

Zero-Trust Local Privacy: All task histories and state stored in local SQLite WAL databases

Operating System Keyring Vault: OS-level encryption for API keys and credentials

Screen Understanding: Low-latency local frame capture and vision-language model integration

Model Context Protocol (MCP): Connects with standard MCP servers for local tools and developer workflows

Human-in-the-Loop Verification: Destructive actions require argument-aware user authorization

Architectural Decisions & Rationales

Tauri v2 + Rust backend rather than Electron

Rationale: Reduces memory footprint by 80% (<120MB idle vs >600MB for Electron) while providing direct POSIX API access and memory safety.

Local-first SQLite storage with OS Keyring encryption

Rationale: Ensures zero telemetry or confidential file contents ever leave the host machine without explicit operator consent.

Technical Trade-offs & Honest Limitations

Desktop vision understanding requires capable local hardware or fast connection to self-hosted/private model endpoints.
Requires OS-level accessibility and screen recording permissions on macOS and Wayland compositors.

Quickstart & Execution

# Clone and run Kyro in development mode

git clone https://github.com/eshanized/Kyro.git

cd Kyro

npm install

npm run tauri dev

Empirical Verification Evidence

Public repository at github.com/eshanized/Kyro, official documentation portal at kyro.tonmoyinfrastructure.org, and AppImage releases built on GitHub Actions CI.

Related Research & Architectural Writing

Journal Essay

Designing Kyro: Local-First Desktop Automation with Tauri v2, Rust, and MCP

How we built Kyro, a local-first autonomous desktop AI agent that bridges screen understanding and real computer use with zero-trust local SQLite privacy and OS Keyring security.