Source note
Show HN: Collaborative context-sharing memory platform for agents and teams
Summary
xysq is a shared memory layer for AI agents and teams. It combines data from workplace tools into isolated, consent-controlled team vaults that agents can query across applications.
Problem
- Team context is scattered across Slack, Drive, Notion, email, and individual workspaces, so agents lack persistent access to decisions and working knowledge.
- Useful context can disappear when employees change roles or leave, increasing repeated work and weakening institutional memory.
- Sharing memory across agents creates privacy, ownership, deletion, and model-training concerns.
Approach
- Connectors ingest files and conversations from existing team tools and organize them into a living knowledge graph.
- Team vaults keep each organization’s memory isolated and expose it to authorized agents through an app, API, SDKs, and reference architectures.
- Consent gates cross-agent access, while encryption protects memory in transit and at rest.
- Users can inspect, export, and delete stored memory, and xysq states that it does not use customer memory to train models.
Results
- The excerpt provides no quantitative results, benchmark dataset, latency measurement, accuracy metric, or baseline comparison.
- The strongest concrete product claims are persistent context across connected AI tools, queryable team knowledge from Slack, Drive, Notion, and other sources, and integration without rebuilding existing agents.
- The product claims user-controlled memory ownership through consent-gated access, export, and deletion controls, but the excerpt provides no independent verification or deployment data.