ViBo is a graph memory service for AI agents and bots. It stores what your agent knows — connected, encrypted, and impossible to copy. Semantic search finds the right facts in milliseconds, not megabytes.
Instead of loading all memory into the prompt, the agent retrieves only the relevant nodes — ranked by lemmatization and semantic embeddings, encrypted at rest.
Every fact becomes a node in the knowledge graph.
Related facts are linked by edges — associations, not just entries.
Semantic search understands «what do I like?», not just exact phrases.
L1/L2/L3 tiers — secrets never leak into the LLM context.
Tested on a real graph of 100K nodes. These are measurements, not estimates.
| Flat files | Vector DB | Built-in memory | ViBo | |
|---|---|---|---|---|
| Token cost | $0.022/req | high | opaque | $0.0003/req |
| Encryption tiers | none | none | none | L1/L2/L3 |
| Portability | .md | cloud-locked | locked | one .web file |
AES-256-GCM with field-level encryption. PBKDF2 600K iterations (OWASP 2023). Secrets never enter the LLM context — verified by tests.
Visible to the agent and LLM. General knowledge, tags, links.
Encrypted with the agent's key. Decrypted only when needed.
API keys, passwords. Never enter the LLM. User-only access.
/activateActivate a license key, bind to a machine.
/statusCheck key status: active / expired / revoked.
/healthService health check — is ViBo up?