Features
Everything your agents remember, made visible.
Hosted memory for AI agents: a live graph, hybrid retrieval, and team roles, connected over MCP. Not a black box — every memory is sourced, editable, and yours.
organization memory · shared ledger
capturing3 memories
Memory lives in the engine; the Cloud surfaces it read-mostly.claude-code
Agents reach memory via a remote MCP endpoint + per-org API key.cursor
agentwiki uses Postgres + pgvector for hybrid memory retrieval.claude-desktop
Twelve capabilities, one shared memory.
Every piece of how stored works, from organization memory to governance. Start anywhere; they all run against the same visible, editable graph.
1.0How it works01
ExploreCapture
- Organization memoryOne shared graph for the whole org: every agent and teammate reads from and writes to the same durable memory.
- Agent memoryEach agent gets its own scoped memory inside the graph: private context that still connects to the shared whole.
- Live memory graphEntities and their relationships appear in real time as your agents write, browsable from the dashboard.
- Entity extractionAs agents write, the engine extracts entities and relationships automatically and adds them to the graph.
02
ExploreRetrieve
03
ExploreGovern
- Provenance and historyEvery memory records who wrote it, when, and from what source, so you can trace an answer back to its origin.
- Visible, editable memoryOpen any memory to inspect it, then correct or delete what an agent got wrong, no black box.
- Verified knowledgeHuman-approved facts become a read-only source of truth agents can rely on without re-verifying.
- Human review queueUncertain or conflicting facts surface for a teammate to confirm before they're treated as verified.
04
ExploreConnect
Know your limits: and your keys.
Track memories, API calls, and embedding tokens against your plan, right from the dashboard. Bring your own OpenAI key — encrypted at rest and scoped to your org.
2.0Pricingusage · against plan ceilings
Stored memories18,420 / 50,000
API calls / month96,500 / 250,000
Embedding tokens1,240,000 / 5,000,000
example usage against real plan ceilings · bring your own OpenAI key
Plan ceilings come straight from the pricing catalog.No surprise overages · example usage, not benchmarks
Give every agent the same memory.
Start with one shared memory layer. Connect the tools your team already uses.