Persistent semantic memory service and MCP adapter for AI agents, powered by PostgreSQL and local Ollama models.
Drift inferred · capture-to-capture
No drift recorded — single capability capture; advisories appear once its surface changes.
tools
-
memory_bootstrap
Load effective rules plus non-rule core recall
-
memory_delete
Delete a memory by UUID
-
memory_get
Fetch a single memory by UUID
-
memory_list
Browse memories by project/category
-
memory_neighbors
List neighbor memories reachable via graph edges
-
memory_recall
Load core memories for a project
-
memory_rules
Load general + project durable rules
-
memory_search
Hybrid semantic search within a project
-
memory_server_version
Return version plus git hash
-
memory_store
Store a new memory
-
memory_update
Update summary/content/tags and re-embed if needed
-
review_queue
List memories tagged review-needed, with optional category filter
-
review_submit
Store a review decision and mark the original reviewed
-
session_finalize
Finalize a shared session into searchable chunks
-
session_log_store
Store a full session transcript for archival/search
-
session_message_append
Append a prompt/response/tool event to a shared session
-
session_start
Create or upsert a normalized shared session
This server hasn't been statically analyzed yet (npm/PyPI artifacts only).