II.
Definition overview
Reference · livedefinition:krate-memory-model
Krate Agent Memory Model (Company Brain) overview
Inspect the raw attributes, linked wiki pages, and inbound or outbound graph edges for definition:krate-memory-model.
Attributes
displayName
Krate Agent Memory Model (Company Brain)
authoredAt
2026-05-10T00:00:00Z
text
The resource kinds implementing Krate's company brain — cross-run
knowledge persistence for AI agents:
AgentMemoryRepository: Org-scoped resource that represents a
knowledge store. Configures the vector index backend, retention
policy, and access control. Each org can have multiple memory
repositories (e.g., per-team, per-project, or org-wide).
AgentMemorySnapshot: An individual knowledge artifact produced
by an agent session. Contains structured metadata (source repo,
PR, session ID, timestamp), a natural-language summary, code
patterns discovered, architectural decisions made, and embedding
vectors for similarity search.
AgentMemoryQuery: A read-only resource used to query the memory
repository. Specifies search terms, filters (time range, repo,
author), similarity threshold, and max results. The aggregated
API server resolves the query against the vector index and
returns matching snapshots.
The memory model enables agents to build cumulative organizational
knowledge: coding patterns, architecture decisions, common error
resolutions, and team conventions persist across independent
agent dispatch runs.
status
canonical
Outgoing edges
applies_to2
- domain:platform-engineering·DomainPlatform Engineering
- domain:software-engineering·DomainSoftware Engineering
supports2
- tool:kubernetes·ToolKubernetes
- skill-area:platform-engineering·SkillArea
Incoming edges
None.