Open-source, self-hosted AI canvas where text, image, video, audio and agent cards share one board, using your own model APIs. Pricing, the evidence and the alternatives, checked Oct 2026.
Source on GitHub under MIT, with 3 stars. Last commit 2026-10-11, so it's actively maintained. Repo checked 2026-10-20.
New & indie Early tool, few public reviews yet. Badge drops once it builds a track record.
Open-source, self-hosted AI canvas where text, image, video, audio and agent cards share one board, using your own model APIs.
Best for
Tinkerers who want a free, self-hosted alternative to paid AI creative canvases and are happy to bring their own model keys or ComfyUI workflows.
Red flags
Very early: v0.0.x development build (v0.0.5, Oct 11 2026) for a single admin, with no promise of data compatibility between versions; you host it with Docker and pay your own model API costs; brand-new project (3 GitHub stars, domain from Oct 2026) with an anonymous maker.
Price
$0
Opening line: “Set up a short-drama project: draft the opening scene, generate two dusk images for the first shot, and have the agent propose a voice-over for approval.”
Agenvas: Open-source, self-hosted AI canvas where text, image, video, audio and agent cards share one board, using your own model APIs. Best for tinkerers who want a free, self-hosted alternative to paid AI creative canvases and are happy to bring their own model keys or ComfyUI workflows. Watch out for: very early: v0.0.x development build (v0.0.5, Oct 11 2026) for a single admin, with no promise of data compatibility between versions; you host it with Docker and pay your own model API costs; brand-new project (3 GitHub stars, domain from Oct 2026) with an anonymous maker. Price: Free and open source (MIT), self-hosted; your model API costs are separate (free tier, checked Oct 2026).
Note: Every asset is a card on an infinite canvas: generate or edit text, images, video and audio per card with queuing, regeneration and version history, and let an agent read the bound context, edit text and place cards, with media proposals running only after you approve them. Bring your own OpenAI-compatible endpoint and media providers (GPT Image, Nano Banana, Seedance, Seed Audio), plus ComfyUI and RunningHub workflows. Runs with one Docker Compose command on macOS, Linux or Windows; no GPU is needed for the app itself. UI in English, Chinese, Russian and Japanese. MIT licensed. The README says it's a development build for single-admin self-hosting without a data compatibility guarantee between versions, so back up before updating.
Free tier
Free
Job in a stack
Images & art, self-hosted AI creative canvas
Cheapest paid plan
Free and open source (MIT), self-hosted; your model API costs are separate
AI pixel art sprites, animations, rotations and tilesets. Best for 2D indie games that need consistent pixel art characters and animations.
Questions people ask
Is Agenvas free?
Free tier. Cheapest paid plan: Free and open source (MIT), self-hosted; your model API costs are separate (checked Oct 2026).
What is Agenvas best for?
Tinkerers who want a free, self-hosted alternative to paid AI creative canvases and are happy to bring their own model keys or ComfyUI workflows.
What are the downsides of Agenvas?
Very early: v0.0.x development build (v0.0.5, Oct 11 2026) for a single admin, with no promise of data compatibility between versions; you host it with Docker and pay your own model API costs; brand-new project (3 GitHub stars, domain from Oct 2026) with an anonymous maker.