Skip to main content
Conare gives your application its own retrieval stack: per-end-user AI memory, a vector + full-text database, and embeddings served from the same model that indexes your data — plus connectors that flow your users’ tools into their memory.

​
Products

AI Memory

Per-end-user persistent memory: save, hybrid search, LLM-synthesized deep recall, and proactive suggestions.

ConareDB

Vector + full-text search engine with isolated namespaces. The substrate under AI Memory, also available standalone.

Embeddings

Document-space vectors from the exact model Conare retrieves with. Stateless, unit-normalized, pinnable.

Integrations

Six first-class connectors plus a catalog of ~200 data sources your end users can connect in one click.

​
Consumption modes

  • HTTP API — your app embeds memory for its end users. Authenticate with an org-owned Integration key (cint_...); every request names an endUserId. Start with the Quickstart or the TypeScript SDK.
  • MCP — coding agents (Claude Code, Codex, Cursor) use Conare directly as an MCP server at https://api.conare.ai/mcp. See MCP tools.

​
Custom models

For teams that want retrieval provably tuned to their own data, Conare trains and maintains custom embedding and reranking models, acceptance-gated on your golden query set. See how the engagement works or book a call.

API Reference

Every endpoint, generated from the same OpenAPI 3.1 spec that production smoke tests verify.