One Memory Layer, Shared.
Capture what your agents learn, verify what is true, and retrieve the same sourced context in every tool.
The shared context layer
A shared memory layer for AI agents. Capture once, verify what is true, and let every connected tool retrieve the same context.
Every memory keeps its source.
Inspect what was remembered, where it came from, and which agent captured it—before that context is used again.
2.0Provenanceagentwiki uses Postgres + pgvector for hybrid memory retrieval.
captured from a working session · read by every connected tool
Provenance
- Written by
- claude-desktop
- Observed
- Jun 26, 2026 · 7:54 PM
- Last edited
- Never edited
- Linked entities
- agentwikiPostgrespgvector
source verified
Find the context behind the query.
Search memories and follow the entities they connect to, so an agent retrieves the relevant context—not an isolated fragment.
3.0Retrieve3 results · ranked by hybrid relevance · same answer in every tool
One connection. Every agent stays current.
Connect each tool to the same memory layer, so Claude Code, Cursor, Windsurf, and Claude Desktop retrieve one shared context.
4.0Connectconnect each tool once · every agent stays current
Same memory layer. Tool-specific setup. Shared retrieval. No exports or re-teaching between sessions.
See the memory layer take shape.
Track what has been remembered, which entities were extracted, and how many agents are actively using the workspace.
5.0ObserveMemories
3
stored across the workspace
Entities
4
extracted from memory
This week
+1
new memories · 7 days
Agents
3
writing to shared memory
Give every agent the context your team already knows.
Start with one memory layer. Connect the tools your team already uses.