other / agent

PenEcho

A canvas app for handwriting, equations and diagrams that works with a built-in agent or with Codex, Claude Code and other MCP clients.

Repository owner penecho

Not rated yet
0 comments

Like 0 Save

Published 11 Oct 2026Listing updated 12 Oct 2026

At a glance

Purpose: A canvas app for handwriting, equations and diagrams that works with a built-in agent or with Codex, Claude Code and other MCP clients.
Runs in: MCP server, Web service
Operating systems: macOS, Windows
Needs: Runtime: Node.js 22.19+ for npm installation; Needs internet: Yes; Needs accounts or API keys beyond your agent: Not stated
Files and data: Reads files: Not stated; Writes files: Yes; Sends data to other services: Yes
Tool type: Agent
Delivery: From GitHub
Version: Latest on the main branch
Access: Code on GitHub
Compatibility: The README lists desktop apps for macOS and Windows, an npm route requiring Node.js 22.19+, and hosted-browser access. The README names Codex, Claude Code, Kimi and Cursor as MCP clients, and lists OpenAI- or Anthropic-compatible APIs and signed-in Codex, Claude Code or Kimi CLIs as model connections.
External costs: According to the README, PenEcho models use account credits, your own model API is billed by your provider, and the CLI route uses your own plan. The README says PenEchoLLM suggestions are free to try every day with no API key. It gives no prices. A commercial licence is available for the code. The providers you connect may charge for usage.
Permissions: The README shows a global npm installation and a source route using npm install. The README says the desktop app updates itself. The README describes model connections using the user's API key, signed-in Codex, Claude Code or Kimi CLI, or hosted PenEcho models. The README says MCP clients can be connected through Auto configure or a generated setup prompt, and agents receive access to canvases the user turns on. The README distinguishes local MCP from a separate signed-in Cloud MCP connection for Cloud canvases. The README says first start offers a six-digit access code or open access on what it calls a trusted network. The top-level file list includes scripts/, whose purpose the README does not describe. The README does not describe general file or shell permissions for the agent.
Usage terms: GNU Affero General Public License v3.0, as stated in the repository. See its LICENSE file for the exact terms. The README says the licence is AGPL-3.0-only, that a commercial licence is available (see COMMERCIAL-LICENSE.md), and that its diagram renderers adapt helpers from another project listed as MIT (see NOTICE).
Interface: desktop

Code on GitHub

penecho/penecho

AGPL-3.0

JavaScript2,480 stars318 forksLast pushed 12 Oct 2026 UTC

Think with AI beyond the chat box. A shared canvas for handwriting, equations, diagrams, and spatial reasoning.

Topics: ai canvas claude codex education handwriting nodejs visual-thinking

Public GitHub data fetched 13 Oct 2026 UTC · 0 days old

Skill folders in this repository (4)

Folders that contain a SKILL.md file, from GitHub's file list on 13 Oct 2026 UTC.

View on GitHub ↗

What it does

The README describes PenEcho as a canvas for handwriting, equations, text and images, with a built-in Agent or external MCP clients such as Codex and Claude Code.

The README describes Smart Assist with PenEchoLLM suggesting actions such as Solve, Typeset, Plot graph, Make diagram and Organize as Note. The README describes Lasso AI for selected canvas content, Notes & Cards in a searchable Library, widget refinement through written feedback, and architecture, sequence and workflow diagrams with SVG or PNG export.

The README gives desktop, npm and hosted-browser routes. The README lists PenEcho models using account credits, the user's OpenAI- or Anthropic-compatible API, and signed-in Codex, Claude Code or Kimi CLIs as model connections.

Who it is for

People who want to work with an AI on a handwriting canvas, and users of Codex, Claude Code or other MCP clients who want their agent to draw on it.

Example tasks

Install with npm, run penecho and open the canvas in a browser.
Add a model in Settings, AI & connections, with your own API key, a signed-in CLI or PenEcho models.
Turn on the current canvas under Settings, MCP service, and ask a connected agent to show an architecture on it.
Pause after writing a handwritten equation and pick a suggested action such as Solve or Typeset.
Circle part of the canvas with Lasso AI to solve, visualize, animate or organize just that content.

Getting started

The README gives three routes. For the desktop app, download it for macOS or Windows from GitHub Releases. For npm, with Node.js 22.19+, run npm install -g penecho, then penecho, and open localhost:3888. Or sign in at the hosted site and install nothing. The README says to choose a six-digit access code or open access on what it calls a trusted network, then add a model in Settings, AI & connections. To connect an agent over MCP, turn on the current canvas in Settings, MCP service, and use Auto configure or copy the setup prompt into your client. The README has the exact commands.

Limitations

According to the README, the npm route needs Node.js 22.19 or newer. The README says first start offers a six-digit access code or open access on what it calls a trusted network. The README says to add a model in Settings and lists three options: your own API key, a signed-in Codex, Claude Code or Kimi CLI, or PenEcho models that use account credits. The README says agents see canvases you turn on. The README instructs users to sign in at the hosted site for hosted models and Cloud canvases. The README names macOS and Windows for the desktop app. It does not describe file, shell or network permissions for the agent. The README says the licence is AGPL-3.0-only, with a commercial licence available. The repository can change at any time; check its README before you install.

Data practices

The README says the desktop app or CLI serves the canvas on your computer and uses your own model API or agent CLI. The README says the hosted site adds hosted models, synced projects and favorites, and can reach your linked computer. The README says PenEchoLLM learns from the suggestions you choose, on your device. The README says Local MCP stays on your computer and Cloud MCP is a separate signed-in connection for Cloud canvases. The README does not describe telemetry or specify what data each model connection sends.