14:45:20
writememory.store — KYC attestation linked
prev 23cd…e4d5 ⛓ hash 690d…2131
Encrypted at rest. Immutable, tamper-evident audit trails. Contradiction detection across every session. Velixar is the memory layer for AI that has to answer to a regulator — not just a user.
Free tier, no card. Or start from your terminal: npm i velixar · pip install velixar
Vector stores retrieve. Memory libraries persist. Neither was built to be audited. Velixar treats every memory as a governed record from the moment it's written.
AES-256-GCM envelope encryption on every memory. Multi-tenant workspace isolation with role-based access — members, admins, and agents each see exactly what policy allows.
Every write, recall, and change is tamper-evident and timestamped. Export the full trail for a regulator, a customer audit, or an internal review — on demand.
Memories link forward and backward in time. When new information conflicts with what the system believed before, Velixar surfaces the contradiction instead of silently overwriting it.
Each node in the lattice above is a record like these: written, tamper-evident to what came before, and exportable the moment someone asks. This is what your compliance team sees.
| Capability | Vector DB | RAG pipeline | Memory library | Velixar |
|---|---|---|---|---|
| Semantic recall | ✓ | ✓ | ✓ | ✓ |
| Persistent across sessions | — | — | ✓ | ✓ |
| Encryption at rest, per-memory | — | — | — | ✓ |
| Immutable, exportable audit trail | — | — | — | ✓ |
| Identity model & belief tracking | — | — | — | ✓ |
| Contradiction detection | — | — | — | ✓ |
| Role-based access for compliance | — | — | — | ✓ |
Open-source SDKs and an MCP server that drops into Claude Desktop, Cursor, Windsurf, or any MCP host. Thin clients, fully public — the governed engine stays server-side.
from velixar import Velixar v = Velixar(api_key="vlx_...") v.store("Client risk profile moved to conservative") results = v.search("risk profile", limit=5) # every call above is already in the audit trail
import Velixar from 'velixar'; const v = new Velixar({ apiKey: 'vlx_...' }); const { id } = await v.store('User prefers dark mode'); const { memories } = await v.search('preferences'); // every call above is already in the audit trail
// Claude Desktop / Cursor — mcpServers config { "mcpServers": { "velixar": { "command": "npx", "args": ["-y", "velixar-mcp-server"], "env": { "VELIXAR_API_KEY": "vlx_..." } } } }
In finance, memory is a compliance requirement. Velixar gives AI systems an immutable, timestamped record of every interaction, recommendation, and rationale — ready for review on demand.
Explore Velixar Finance →Every learner gets a persistent, private profile — what they know, where they struggle, how they learn best — with access controls a district can approve.
Explore Velixar Education →A proper memory layer — identity, temporal chains, cognitive signals — in three lines of code. Plugs into any LLM or agent framework.
Explore Velixar Developer →Talk to us about governed memory for your regulated workload — or mint an API key and make your first audited call in under a minute.
On the Record
Bring the right books, check the card, truly withdraw the withdrawn. GateMem scores all three at once — and Velixar is the only librarian that passes the whole test.
Read →Long context beat Velixar on GateMem, and we published it. The real question isn't benchmark scale — it's what happens to cost, latency, and governance as history grows.
Read →We benchmarked Velixar on HaluMem expecting a bigger extraction model to win. It didn't — and our obvious fix for the weakest category made things worse. What component isolation taught us.
Read →