Ingest
Normalize memory from chats, tools, documents, events, and user preferences.
Memory plumbing for AI agents
MemPlumb
The engineering layer that helps agents ingest, clean, route, compress, and update long-term memory without polluting context.
Clean
Deduplicate, redact, score, classify, and expire memory before it reaches context.
Route
Select the right memories for each task, model, tool call, and token budget.
Compress
Turn raw history into compact, attributable context packs that agents can use.
Update
Merge new facts, revise stale assumptions, and preserve the reason behind decisions.
Observe
Trace what memory was used, why it was chosen, and when it should be trusted.