Willder
AI agents coordinate. People collaborate. One memory.
11 followers
AI agents coordinate. People collaborate. One memory.
11 followers
Every AI tool I used forgot everything the second a session ended. So I built Willder: one shared memory graph that a team works through together, with real access control gating every read and write. Each person can run multiple customizable AI agents on top of it. Agents can coordinate through the graph without sharing a conversation. Nothing an agent drafts ships without your approval. Built solo. Would love your feedback, happy to talk through the architecture or the access-control model.
This is the 2nd launch from Willder. View more
Willder
Launching today
A context layer that syncs your apps, agents and teams. Each agent gets its own scoped, revocable token; what one agent writes, the others read on their next call; connectors bring your apps in under the same rules (measured on GitHub so far). The engine is MIT and self-hostable.
Free
Launch Team / Built With
AWS Startup AdvisorPersonalized AWS guidance that meets you where you’re already building
Promoted
Hi Product Hunt, I'm Tri, founder of Willder.
I run my company solo, with AI agents across product, sales, marketing, finance and legal. 20 of them, 24/7, and they build and use Willder. Before Willder they overwrote each other, worked from stale information, and had access they shouldn't. In a May 2026 Akeyless survey of 400 IT and security leaders, two-thirds of organizations using AI agents suspected those agents had already accessed data beyond their scope.
Willder is a context layer that syncs your apps, agents and teams:
What one agent saves, the others read on their next call, with who wrote it.
Fix a fact once and every agent reads the new version on its next call.
Each agent only sees what its own scoped, expiring token allows, and a revoked token is refused on its next call.
Every fact shows who wrote it, where it came from, what it replaced and which agents read it.
GitHub, Drive, Slack and Figma land in scopes you choose, under the same rules (measured on GitHub so far).
The engine is open source (MIT): https://github.com/willderai/willder-core. The hosted version is at willder.ai, and Claude Code and Codex connect over MCP.
AI is the wilder where you're the wielder. What's the most annoying thing your encounter running multiple agents, especially with others?