Gnosis: 3D, local-first visualization for multi-agent coding
Gnosis, developed by DOMCHURCH, is a local-first terminal coding agent that visualizes multi-agent AI workflows for developers and researchers. The app replaces scrolling terminal output with an interactive Three.js 3D office that shows agents moving between Planning, Coding, and Verification zones in real time. It supports mid-session model switching, aggressive prompt caching, Tree-sitter code indexing, and MCP tool integrations. This combination targets software teams who need transparent, local control over agent-driven coding tasks.
What tasks can you actually use it for?
The tool maps multi-agent activity into a spatial dashboard so users can monitor coordination and sub-agent behavior visually. In the 3D office agents occupy zones such as Coordinator, Planning, Coding, and Application, and the interface includes a local PTY terminal stream and a file diff viewer. Use cases include monitoring parallel code-generation tasks, tracking verification jobs, and overseeing scoped sub-agents during complex builds.
How reliable is session continuity and context retention?
Gnosis preserves conversation and tool state when switching models mid-session via OpenRouter integration, which supports model hot-swapping without losing context. The app also applies aggressive prompt caching to reduce API round trips, a practice the developer reports can cut external calls by over 10x. Tree-sitter based code indexing supplies deeper local context for code-oriented prompts, improving the relevance of agent outputs tied to repository content.
What inputs and environment requirements does it impose?
This is a terminal-first application built on Node.js and TypeScript and it requires an environment that supports npm. The app functions as an MCP host and integrates with MCP tools, so users must configure MCP-compatible services to unlock tool integrations such as Playwright or Chrome DevTools. Automated test suites have verified operation on Windows, indicating cross-platform development expectations but a developer-focused install path.
Does it fit existing developer workflows and protect local data?
The app is local-first and emphasizes local PTY control and Obsidian integration for persistent memory, which suits teams that prefer on-machine data handling over cloud storage. MCP support connects agents to existing developer tooling, and mobile or LAN control offers remote session management. The developer community response highlights the value of a spatial, agentic UI for making complex workflows observable during active development.
A targeted tool for developers who need visible, local agent control
The tool is a practical option for software developers and AI engineers who prioritize transparent, on-machine management of agentic workflows, especially when preserving session context matters. Expect a technical setup and a learning curve for mapping operations into the 3D office metaphors. For teams that accept that upfront effort, the tool rewards clearer oversight of parallel agent activity and local data handling.
