PenEcho offers a shared canvas for freeform input (handwriting, equations, diagrams), enabling AI to interact directly with spatial context. It solves the rigidity of traditional chat and diagram tools, allowing users to "think with AI beyond the chat box" by receiving AI drafts, hints, or explanations directly on the canvas, fostering integrated creative and problem-solving workflows.

How It Works

User input on a sparse, zoomable canvas is sent to a PenEcho server, which prompts a configured LLM executor (API, Codex CLI, or Claude CLI). The AI generates an editable draft response on the canvas, separate from confirmed ink, until accepted or discarded. This approach leverages spatial relationships for iterative AI refinement within the user's work.

Quick Start & Requirements

• Install: npm install -g penecho
• Prerequisites: Node.js 18.17+ and an LLM backend (API key or authenticated Codex/Claude CLI).
• Configuration: Run penecho configure for interactive LLM setup.
• Run: Execute penecho after configuration.

Highlighted Details

• Per-request Reasoning Control: Six levels (None to Max) adjust AI quality/speed.
• High-Resolution Export: Generates crisp PNGs of confirmed ink with margins.
• Typed Text & LaTeX: Supports Markdown and inline LaTeX in movable editors.
• Sparse Canvas: Efficiently manages large logical canvases by allocating tiles only where content exists.

Maintenance & Community

Actively developed in the open by a small community. Support and discussion via Discord, GitHub Discussions, and Issues. Contributions welcome for model testing (esp. Gemini), bug reporting, and code.

Licensing & Compatibility

Licensed under GNU AGPL v3.0. Commercial use permitted under AGPL (requires network-provided modifications to be open-sourced). An alternative commercial license is available for non-AGPL compliant products.

Limitations & Caveats

Described as "young," indicating ongoing development. Google Gemini models are untested. Security requires careful management: plaintext API credentials locally, and CLI modes should not be exposed to untrusted networks. Public API deployments need robust external security.