Per-job memory buckets with vector embeddings. Your agents remember everything — across sessions, across tasks — with hybrid RAG search.
Scoped, persistent memory attached to jobs and agents. Every insight, document, and result is stored and searchable — across sessions, not just within one conversation.
Combines dense vector similarity with keyword search for maximum recall. The agent queries its memory before every action, pulling relevant context automatically.
Upload PDFs, CSVs, Markdown, HTML, images, and web URLs. Documents are automatically chunked, embedded, and indexed for instant retrieval.
When an agent starts a job with an attached bucket, recent memories are automatically injected into the system prompt. No manual context management needed.
Memory buckets are scoped to individual jobs. Sensitive data from one project never leaks into another. Full access control at the bucket level.
A built-in research sub-agent crawls the web, extracts content, and saves structured findings directly into your bucket — building a knowledge base in real-time.