Production agent memory—continuous across sessions, predictable costs
Layered wake-up loads core context at session start and expands on user questions—replacing full knowledge-base dumps every chat. Built for support, engineering copilots, and internal assistants: cross-session continuity, stabler Q&A, controllable token cost.
For support bots, copilots, and internal assistants — MCP · REST API · enterprise-ready
Agent Memory
Public benchmarks you can reproduce
ENGRA-KB-v2 (550 docs, 400 queries)—structured atoms vs BM25 and fixed-size chunk RAG
76.0% Recall@10 · 0.301 MRR · 550/400 Docs / queries — ENGRA-KB-v2 (550/400)
Agent Memory · synthetic enterprise corpus · same embedder · retrieval-only comparison
View full Memory benchmark results →Agent Memory
The capability that takes agents from PoC to production: structured memory instead of arbitrary chunks; layered wake-up instead of full-corpus search every session. Fewer wrong answers, less user repetition, predictable token spend
Agent Memory
Long-term memory production agents remember—with predictable token spend
- →Multi-team governance: Library / Scope / Topic memory boundaries
- →Layered wake-up: core context at start, details on demand—predictable token cost
- →Traceable links: path backlinks for audit and knowledge evolution
- →Published benchmark: ENGRA-KB-v2 Recall@10 75.3% (+20pp vs chunk RAG)
- →Production access: Memory MCP + REST API, quotas and SLA
Learn more →
Pick your integration path
One atom model everywhere: IDE agents via MCP, services via REST, SDK on the roadmap
Cursor / Claude / Copilot / Cline & more
MCP + Skills for IDE agents
Streamable HTTP MCP tools plus SKILL.md playbooks—wake up, search, and persist atoms in about five minutes
/api/v1/memory/mcp · Streamable HTTPSKILL.md · 20+ agentsCustom backends / orchestration
REST API
Full Library / Scope / Topic / Atom HTTP surface with built-in auth and quotas
/api/v1/memory/*REST docs →
In-process npm embed—same memory model as remote access