Memory
Sovereign AI memory infrastructure
Chordian unifies your organisation's knowledge into a typed, temporal memory that answers with citations — on our infrastructure or entirely on yours.
curl https://api.chordian.ai/api/v1/memory/search \
-H "x-api-key: $CHORDIAN_API_KEY" \
-H "content-type: application/json" \
-d '{"query": "Status of the Acme expansion?"}'Enterprise
What Chordian is, how the memory model works, and how to run it — including on your own infrastructure.
Start with the overviewPlatform API
Workspaces, memories, credentials, spaces, webhooks, usage and audit. The control plane for your account.
Browse the referenceMemory API
Ingest content, run hybrid retrieval, traverse the knowledge graph, and keep it accurate over time.
Browse the referenceWhat makes it different
Six capabilities that separate a memory platform from a vector database with a chat box on top.
Hybrid retrieval
Vectors, graph traversal, temporal recall and cache run in parallel, are reranked, then synthesised with inline citations. Sources arrive before the answer.
Read moreA knowledge graph, not a chunk pile
Typed entities and relationships with bi-temporal facts — what was true, and when you learned it. Duplicates across sources resolve to one entity.
Read moreMemory that defends its accuracy
When two sources disagree, Chordian surfaces the conflict instead of letting the last write win — then scores its own health and compiles what it knows.
Read moreA perimeter around your content
Secrets and PII are detected, classified and masked at the value level before anything is embedded, stored, or shown to a model.
Read moreYour memory, your metal
Run a memory on your own server or cloud account. Content, embeddings, graph and usage stay on your box — Chordian stores none of it.
Read moreAgent-native access
MCP endpoints with OAuth, per-memory credentials and scoped keys — so the memory reaches the IDE and the assistant, not just the browser.
Read moreSecurity & compliance
Defence in depth — from post-quantum transport to data sovereignty. See the full security overview.
Post-quantum cryptography
Hybrid post-quantum TLS key exchange (X25519 + ML-KEM-768, FIPS 203) at the edge — hardening transport against harvest-now, decrypt-later.
FlagshipGDPR & data residency
Regional Secure Dedicated memories keep content — and inference — in the region you choose.
Tenant isolation
A tenant scope on every read and write, fail-closed and verified in CI on every change.
Mutual TLS to your boxes
Every call to a customer box presents a client certificate from a private CA — over HTTP and WebSocket alike.
Secrets & Vault
Credentials live in Vault; boxes hold their own key. Secrets are never logged or shipped to a model.
Data sovereignty
Run a memory on your own infrastructure — content, embeddings, graph and usage never reach Chordian.
Audit & RBAC
Five roles, fifty-plus permissions, and an immutable, credential-free audit trail on every privileged action.
Data perimeter (DLP)
Secrets and PII detected, classified and masked at the value level before storage or any model call.
Built to be read by machines too
Every page on this site is available as plain markdown, and /llms.txt indexes the whole thing. If you are building an agent on Chordian, point it here.
curl https://docs.chordian.ai/llms.txt
curl https://docs.chordian.ai/enterprise/quickstart.md