copepodCopepod stores versioned, tenant-scoped memory and turns authorised evidence into compact, cited context for every agent.
Published LoCoMo result92.34% accuracy with 7.1% fewer evaluation tokens than our recorded baselineAgents, people, tools and documents can contribute partial or conflicting information. Copepod stores each item with its tenant, source and version history.
Model-assisted enrichment can add entities, relationships, topics and other useful structure alongside stored memories. Context building is separate: Copepod retrieves source-backed evidence, then assembles a private, versioned result without asking another model to rewrite it.
Combine semantic, keyword, graph and temporal signals, with links back to the memories that support each result.
Copepod queues simultaneous updates, preserves conflict evidence and can use recency rules or human review before the current state changes.
AI enrichment can extract entities, relationships, temporal evidence, sensitivity and PII metadata when the workspace plan and policy allow it.
Tenant-scoped access, agent-bound keys and shared or private visibility keep memory within authorised boundaries.
Create private, revisioned context capsules from authorised evidence, with citations, source fingerprints and a bounded output budget.
Give compatible clients controlled access to memory, search and context tools without building a separate memory layer.
Writes are accepted into a queue and consolidated after processing. Copepod records conflicting evidence, can apply recency rules, or hold ambiguous changes for human review.
Accept each write with tenant and agent scope.
Group updates that refer to the same memory.
Use recency rules or defer to human review.
Publish the current version and retain its sources.
When enabled, AI enrichment extracts entities, relationships, temporal evidence, sensitivity and PII metadata. The original memory and source history remain available.
Copepod-managed AI enrichment routes each task to an appropriate model. Each plan includes a fixed monthly allowance, so spending stays predictable.
For individual builders evaluating Copepod.
For small teams launching their first agent workflow.
For production teams running several agent workflows.
For organisations that need custom capacity, security controls or support.
Connect through MCP or the API, keep sources and versions, and control what becomes shared.