penecho
Think with AI beyond the chat box. A shared canvas for handwriting, equations, diagrams, and spatial reasoning.
What is it?
What it is
A shared AI-powered canvas for handwriting, equations, diagrams, and spatial context, allowing users to interact with AI beyond a standard chat box.
Why it exists
To enable visual thinking where users can work through problems using natural drawing, lasso selections, and interactive widgets without translating every step into text.
Who should use it
Users who want a collaborative, AI-powered canvas for handwriting, equations, diagrams, and visual thinking—such as educators, students, designers, developers, and researchers who value interactive on‑canvas AI assistance.
Who should avoid it
Individuals or organizations that require strict privacy controls, cannot expose API keys or CLI credentials, or need a fully hardened production deployment without local CLI execution.
How it works
A quick walkthrough in plain English
How penecho works in 4 steps
Step 1 of 4
Something triggers the flow
A schedule, webhook, or manual click tells penecho to start.
Features
Advantages
- Spatial context awareness for AI conversations
- Low latency through optimized canvas cropping
- Flexible deployment (local CLI or remote API)
- Extensible plugin architecture with sandboxed HTML
- Supports high-quality reasoning models like Claude Opus and Kimi K3
Disadvantages
- High-effort reasoning requests can incur significant token costs
- CLI modes (Codex/Claude) require local installation and authentication
- Complexity in managing model-specific effort levels
- AGPL-3.0 license requirements for network-distributed modifications
Installation
native
npm install -g penecho penecho configure penecho
FAQ
How do I install and launch PenEcho on my machine?
Install Node.js 18.17+ and run `npm install -g penecho`. Then configure with `penecho configure` to set your LLM source and settings. Finally start the server with `penecho`. The web UI will be available at http://localhost:3888 (or the LAN URLs printed on startup).
What options are available for selecting an LLM source and how do I test them?
In the configuration menu choose `LLM source` → `Claude CLI`, `Codex CLI`, or `API`. Each option lets you pick a model, effort level, and for API mode enter the endpoint, model, and key. After selecting, click `Test & Save`. Codex CLI checks the executable and model list without consuming tokens; Claude CLI and API send a small real request to verify connectivity. The configuration is saved regardless of test outcome.
How can I use the canvas to get AI responses directly on my drawings?
Draw a question, equation, diagram, or any ink on the canvas and pause. PenEcho captures the relevant crop, sends it to the configured LLM, and returns a movable draft next to your ink. You can drag, resize, copy, or accept the draft. Use the toolbar’s `Reasoning` menu to adjust per‑request effort, and choose modes like Arcane, Sci‑fi, Research, or Studio to match your problem type.
Can I run PenEcho from the source repository instead of the global install?
Yes. Clone the repo, run `npm install` to install dependencies, then start with `npm start`. The same interactive configuration will launch if needed. You can also create a global command link with `npm link` and then use `penecho` as usual.
What should I know about token usage and cost when using PenEcho?
Typical output token usage depends on effort: ~1,000 for low, ~3,000 for medium, and 5,000–8,000 for xhigh/max. Costs vary by provider and model; for example, a low‑effort request with gpt‑5.6‑sol might cost around $0.08. Check the provider’s pricing page for current rates and remember that Codex and Claude CLI use your account’s limits, not API keys.
Featured in Videos
YouTube tutorials and walkthroughs for penecho
Alternatives
Similar projects ranked by category, topics, and text overlap.
