Cognitive graph and temporal memory
Store entities, relationships, and temporal timelines with automated entity resolution: the structured half of ClawDB's hybrid retrieval engine.
A plain vector index can tell you that two pieces of text are similar. It can't tell you that "the customer's plan" changed from Starter to Pro last month, or that two support tickets are actually about the same underlying account. The knowledge graph is what lets an agent reason over relationships and time, not just word similarity.
Time-aware facts
Tracks how facts change over time. Know when a user changed preferences or when project specs updated, and what the old value used to be.
Auto-deduplication
Automatically merges duplicate entities like "John Smith" and "J. Smith" into a unified graph node, so lookups don't miss half the facts.
Contradiction pruning
Detects conflicting memories and synthesizes consolidated knowledge summaries for LLM prompts, instead of feeding an agent two versions of the truth.
Traverse, not just search
Ask "what does this entity connect to" as a graph traversal, something a plain vector index has no concept of at all.
Hybrid retrieval beats vector-only search
Pure vector similarity finds what's semantically close. It can't tell you what changed, what contradicts what, or how two facts relate to each other. ClawDB pairs its HNSW vector index with a graph layer so agents can reason over structure, not just proximity. In practice, that means fewer stale answers and fewer moments where an agent confidently repeats something that's no longer true.
Common questions
No. Entities and relationships get created as your agent writes memory. There's no separate schema migration step, though you can constrain relation types if you want stricter data.
Give your AI agents permanent memory today.
Deploy in under 2 minutes. Free tier included forever.
Launch Studio Free →