Retrieval
Project-scoped, related-project and global retrieval with lexical, semantic and hybrid modes.
Shared memory layer for agents
Memory system for real agent workflows
It is built for agent loops that need more than a transient prompt buffer: search, relevant context, imports, review queues, quality gates, lifecycle cleanup, compaction and operator visibility.
Project-scoped, related-project and global retrieval with lexical, semantic and hybrid modes.
Quality scoring, semantic duplicate detection, operator review actions and lifecycle maintenance.
Structured project import, reimport, conflict detection and doc-aware ingestion for real repositories.
Runtime self-checks, metrics, review queues, task logs, compaction and embedded admin console.
The public surface is intentionally small. Production API lives on a dedicated subdomain. Admin runtime is kept separate and is not linked from this page.
curl -sS 'https://api.memlayer.ru/health'
For a lightweight read-only smoke check there is a dedicated endpoint. Use your own API key and keep it out of shell history if needed.
curl -sS 'https://api.memlayer.ru/admin/runtime/self-check?search_query=architecture&limit=5' \
-H 'Authorization: Bearer YOUR_API_KEY'
MemLayer does not only store knowledge. It also checks whether the stored knowledge remains useful: placeholder rejection, semantic duplicate signals, review queues, conflict resolution and compaction are part of the runtime.
Project scans can bootstrap memory from real repositories, including hidden docs, handoff files and structured artifacts. The system can then retrieve context by project, related projects or globally.
curl -sS 'https://api.memlayer.ru/memory/search?query=architecture%20decision&mode=hybrid&limit=8' \
-H 'Authorization: Bearer YOUR_API_KEY'