Kanwas: open-source canvas where humans and AI agents share the same product context, every doc is a plain .md file with git history, no vendor lock-in
Launched at #1 on Product Hunt in May 2026 as "the team's brain." The pitch is a familiar one — yet another shared workspace promising to replace Notion plus Obsidian plus Claude chats. What makes Kanwas worth a second look is not the canvas itself but what sits underneath it: every board, note and decision is a plain markdown file with git-backed version history, so the team owns the files even if they walk away from the product tomorrow.
The compounding context layer is where the actual leverage is. ChatGPT and Claude write fluent but generic strategy drafts because they never see the full picture — user calls, past positioning, investor feedback, half-finished bets. Kanwas keeps all of that on one board and feeds it back into every next agent run, so the model starts loaded with the team's reality instead of a cold prompt. The interesting bet is treating context as a compounding asset rather than a passive archive.
Model-agnostic by design — Claude, GPT, Gemini, whatever stack the team prefers — with a terminal-grade agent that does not force everyone into a terminal, real-time collaboration, and 1,000+ connectors plus a CLI to pull in data from tools already in use.
The honest read: the "shared brain" framing is overplayed by half the launches this year, and most founders do not need yet another canvas. What is genuinely useful here is the git-backed .md substrate plus the agentic layer on top — it is closer to a "Cursor for strategy work" than to a Notion clone, and that is a much rarer thing in the market.
One reference on the lander: Samuel Beek (Schematik) ran user calls, investor conversations and positioning through Kanwas, iterated the pitch deck inside it, and closed a €4.6M pre-seed a week later. One data point, but the workflow it describes — context → deck → iteration → close — is the exact loop early teams run by hand for weeks.
Free to start, browser-based, open-source on GitHub.
https://kanwas.ai/
Launched at #1 on Product Hunt in May 2026 as "the team's brain." The pitch is a familiar one — yet another shared workspace promising to replace Notion plus Obsidian plus Claude chats. What makes Kanwas worth a second look is not the canvas itself but what sits underneath it: every board, note and decision is a plain markdown file with git-backed version history, so the team owns the files even if they walk away from the product tomorrow.
The compounding context layer is where the actual leverage is. ChatGPT and Claude write fluent but generic strategy drafts because they never see the full picture — user calls, past positioning, investor feedback, half-finished bets. Kanwas keeps all of that on one board and feeds it back into every next agent run, so the model starts loaded with the team's reality instead of a cold prompt. The interesting bet is treating context as a compounding asset rather than a passive archive.
Model-agnostic by design — Claude, GPT, Gemini, whatever stack the team prefers — with a terminal-grade agent that does not force everyone into a terminal, real-time collaboration, and 1,000+ connectors plus a CLI to pull in data from tools already in use.
The honest read: the "shared brain" framing is overplayed by half the launches this year, and most founders do not need yet another canvas. What is genuinely useful here is the git-backed .md substrate plus the agentic layer on top — it is closer to a "Cursor for strategy work" than to a Notion clone, and that is a much rarer thing in the market.
One reference on the lander: Samuel Beek (Schematik) ran user calls, investor conversations and positioning through Kanwas, iterated the pitch deck inside it, and closed a €4.6M pre-seed a week later. One data point, but the workflow it describes — context → deck → iteration → close — is the exact loop early teams run by hand for weeks.
Free to start, browser-based, open-source on GitHub.
https://kanwas.ai/