High-performance AI infrastructure for general use, research, code intelligence, and multi-provider API access.

Brainwires Studio is a full-stack AI platform for research, coding, and agent automation, centered on three primary apps: the Next.js web experience, the Brainwires CLI/TUI, and the Thalora headless browser. The web app, CLI, and Thalora ship as binary-only releases; the public repos focus on supporting engines, agent tooling, and selected infrastructure forks used in production.

A Next.js 16 Progressive Web App powering brainwires.studio.

• Multi-Provider AI Chat — OpenAI, Anthropic Claude, Google Gemini, and Groq
• Workspace Management — Team workspaces with role-based access control
• Private RAG — Client-side document search using local embeddings
• MCP Integration — Model Context Protocol servers for extended AI capabilities
• Internationalization — Full i18n support with i18next
• PWA & Offline — Service worker caching, installable app

Next.js 16 | React 19 | TypeScript | Supabase | Tailwind CSS | shadcn/ui

The studio integrates several high-performance Rust modules compiled to WebAssembly:

A full-featured pure Rust headless browser designed for AI agents. Binary-only closed source.

• Chrome 131 compatibility with real JavaScript execution
• Modern Web APIs (Fetch, WebRTC, WebAssembly, Service Workers, WebGL)
• AI Memory Heap for persistent context across sessions
• 17+ MCP tools plus Chrome DevTools Protocol integration
• Native and WASM builds, single-binary deployment

A comprehensive mathematical and scientific computing library providing 400+ operations through a unified 10-tool API.

• Custom Computer Algebra System (CAS) with no Python dependencies
• 1,685 tests with 100% pass rate and 83%+ code coverage
• Covers 24+ mathematical domains: calculus, linear algebra, tensor calculus, differential equations, quantum mechanics
• Multiple interfaces: Native Rust API, JSON/MCP Server, WebAssembly

Tools: Solve | Differentiate | Integrate | Analyze | Simulate | Compute | Transform | FieldTheory | Sample | Optimize

Universal Programmatic Tool Calling implementation for any LLM.

• 37-98% token reduction via Rhai script orchestration
• Model-agnostic: works with Claude, OpenAI, Gemini, local models
• Sandboxed execution with configurable safety limits
• Native Rust + WASM (browser/Node.js) targets

Client-side LLM inference engine with hybrid local/cloud support.

• WebGPU acceleration (80% native performance)
• Intelligent fallback: WebGPU → WASM → Cloud
• Smart storage with File System API + Cache API
• Curated model catalog (1B-8B params) with q4/q8 quantization

High-performance client-side vector database.

• WASM-accelerated with near-native performance
• HNSW index for approximate nearest neighbor search
• Automatic IndexedDB persistence
• Cosine similarity, Euclidean distance, dot product

Brainwires Extension — Client-side compute for brainwires.studio.

• Offload heavy computation — Run tools locally instead of paying server costs
• Run AI models locally with WebGPU acceleration
• 400+ math operations via Computational Engine WASM
• Web scraping and content extraction via Thalora
• Chrome (Manifest V3) and Firefox support
• Zero telemetry — all compute stays on your machine

An AI-powered agentic CLI tool for autonomous coding assistance. Binary-only closed source.

• Multi-agent architecture with orchestrator/worker roles
• Rich tool system: file ops, bash, git, web operations
• MCP client plus MCP server mode
• Interactive, single-shot, batch, and TUI chat modes
• Multiple output formats and quiet mode for scripting
• Planning mode, task management, plan branching/templates
• Infinite context memory with entity graphs and semantic search
• Remote control via Brainwires Studio web interface

Below is the full list of public repos in the Brainwires org, with a short note on what they are and why they exist.

What: Standalone computational engine with 400+ mathematical operations. Why: Core math engine powering Brainwires tools, open for reuse and audits.

What: Rust MCP server for RAG over large codebases. Why: Provides semantic search + retrieval infrastructure for assistants.

What: Programmatic Tool Calling (PTC) implementation for any LLM. Why: Cuts token usage and enforces structured tool routing across models.

What: Rust MCP server exposing SSH/SCP workflows as tools. Why: Safe, auditable remote operations for agents and deployments.

What: MCP server that can restart Claude Code during MCP updates. Why: Solves self-restart deadlocks when an MCP updates itself.

What: Interactive TUI primitives for Ratatui. Why: Powers the Brainwires CLI/TUI UX with reusable components.

What: GPU-accelerated React components built on Skia/CanvasKit. Why: High-performance data visualization for web UIs.

What: MCP server for self-hosted Supabase deployments. Why: Bridges agent tooling with our internal Supabase stacks.

What: Supabase Postgres fork with Spock multi-master replication support. Why: Used for pgEdge deployments (Spock = multi-master replication).

What: Supabase CLI fork (migrations, local dev, edge functions, backups). Why: Maintained to support pgEdge/Spock workflows and custom patches.

What: Supabase platform repo (dashboard, infra, tooling). Why: Source-of-truth fork for pgEdge-specific integration work.

What: Next.js PWA plugin fork. Why: Maintains compatibility fixes and build tweaks for our web app.

• Performance First — All core projects are written in Rust for maximum speed and safety
• AI-Native — Built for seamless integration with AI assistants via MCP
• Full-Stack Ownership — We build the browser, the compute engine, the CLI, and the web app

• Website: brainwires.studio
• YouTube: @BrainwiresCortex
• Business Inquiries: support@brainwires.net

_{Building tools that think alongside you.}