Voicetree is a local‑first spatial markdown editor that integrates AI agents into a graph‑based workspace, allowing users to navigate and recall context without manual reminders. It runs existing agents (e.g., Claude Code, Gemini CLI) in embedded terminals and keeps all notes as plain Markdown files on the user’s machine.
Funding
Funding not disclosed
Founders
Product
Problem
Users of AI coding assistants and knowledge bases must manually manage context, switch between multiple terminal windows, and constantly remind agents of relevant information, leading to fragmented workflows and lost productivity.
Solution
Voicetree provides a local‑first, graph‑based markdown editor that embeds existing AI agents (e.g., Claude Code, Gemini CLI) in terminal panes within a spatial workspace. The graph stores notes as plain Markdown files on the user's machine and automatically supplies relevant context to agents, eliminating manual reminders. Users can navigate the workspace visually, zooming into specific nodes while preserving the state of each agent session. The system supports recursive task decomposition and maintains an infinite, zero‑rotating context, ensuring that long sessions remain focused and coherent. AI features are optional and can be disabled, allowing the tool to function as a pure spatial editor.
Target Audience
Primary users are developers, researchers, and knowledge workers who rely on AI coding assistants and need a unified, context‑aware environment for managing notes and agent interactions.
Features
- Graph view that visualizes notes and agent terminals, enabling spatial navigation and instant recall of prior work
- Embedded terminals run any locally installed AI agent (Claude Code, Codex, Gemini CLI, OpenCode, OpenClaw, etc.) without requiring plugins
- Automatic shared memory: agents read relevant context directly from the markdown graph, removing the need for manual context passing
- Supports recursive decomposition of tasks, allowing complex problems to be broken into focused sub‑sessions
- Infinite context graph with zero context rot, preserving the full history of notes while providing agents with concise, relevant excerpts
- Local‑first architecture stores all data as plain Markdown on the user's device; AI calls are made only when explicitly invoked
- Optional AI layer can be turned off, letting users use Voicetree solely as a spatial markdown editor