Looking at the source material against the current landing page, I can see several areas where the page has drifted from the current state of Sidespace:
- Hoshi engine: The page says "AI assistant powered by Gemini" but doesn't mention the dual-engine architecture with Legacy Rust fallback
- Missing Squad View details: The page mentions Squad View but doesn't explain the multi-agent cockpit concept properly
- Agent count and roles: Some agent descriptions are incomplete or don't match current implementation
- MCP tools: Says "56 MCP tools" but this needs verification against current state
- Architecture details: Missing key technical details about the brain adapter system
layout: home hero: name: Sidespace text: AI-Powered Development Workspace tagline: The terminal is the engine. Sidespace is the cockpit. actions: - theme: brand text: Get Started link: /workspace/ - theme: alt text: Architecture link: /stack/ features:
- title: Hoshi Brain details: Dual-engine AI assistant (Gemini Pro + Legacy Rust fallback). Manages projects, coordinates agents, executes tasks through MCP tools.
- title: Squad View details: Multi-agent terminal cockpit. Human quarterback coordinates multiple CLI agents with constellation names, real-time monitoring, and persistent sessions.
- title: Memory System details: Vector-powered knowledge base with automated research pipeline. Context persists across sessions — every conversation builds on the last.
Why Sidespace Exists
The terminal has become the center of human/AI collaboration for getting real work done. But raw terminal sessions lack the organizational layer that makes development productive: persistent context, project structure, task tracking, and knowledge management.
Sidespace solves this by wrapping powerful AI coding tools like Claude Code in a native desktop shell that captures, organizes, and resurfaces context across sessions. When your terminal closes, the insights don't evaporate — they become part of your project's persistent knowledge base.
The Architecture
Sidespace is a Tauri 2.0 desktop application with React frontend, Rust backend, and Supabase database. The experience centers around tabbed workspaces: a Hub dashboard for projects and system status, individual Project tabs with kanban boards and documents, and Squad View for multi-agent terminal coordination. A persistent right panel provides instant access to Hoshi, the AI assistant, from anywhere in the app.
The Agent Ecosystem
Sidespace coordinates four specialized agents, each optimized for specific tasks:
| Agent | Role | Model | Interface |
|---|---|---|---|
| Hoshi | Primary assistant and orchestrator. Project management, tool execution, context synthesis. | Gemini 3.1 Pro (Legacy Rust fallback) | In-app chat panel |
| Claude Code | Hands-on development agent. Code writing, debugging, file operations in live terminal. | Claude 3.5 Sonnet (via CLI) | Squad View terminals |
| Umbra | Research and discovery. Scans external sources, surfaces relevant findings. | Claude Haiku | Background pipeline |
| Kosmos | Memory operations. Deduplication, labeling, knowledge base grooming. | Multiple models | Scheduled edge functions |
All agents share a unified tool surface through the Sidespace MCP server — 56 tools for reading and writing projects, tasks, memories, and documents. This means a task created by Claude Code in a terminal is immediately visible to Hoshi in the chat panel, and research findings from Umbra automatically surface in project context.
Squad View: Multi-Agent Coordination
Squad View transforms the traditional single-terminal experience into a multi-agent cockpit. You can run multiple CLI agents simultaneously — each gets a constellation name (Orion, Lyra, Vela) for identity across session logs and the memory system.
The interface provides three zones: a task sidebar for project overview, a focused terminal in the center, and thumbnail terminals at the bottom for quick agent switching. The human acts as quarterback, assigning tasks and monitoring progress across the constellation.
The Memory Advantage
Every interaction feeds into a vector-powered knowledge base. Umbra continuously researches your projects, finding relevant resources and feeding discoveries to Hoshi. Kosmos maintains the knowledge base through automated grooming, deduplication, and labeling. Context builds over time instead of resetting with each session.
This creates a development environment that learns and evolves with your projects — where starting a new session means picking up where you left off, not starting from scratch.
Current State
Sidespace v0.1.0 is in active development and serves as its own primary development environment. The workspace shell, project management, memory pipeline, Squad View, and full agent ecosystem are operational. We're building the future of AI-assisted development by using it to build itself.