# Conare

> Custom retrieval infrastructure — connect, embed, store, retrieve, fine-tune.

## Docs

- [Connect a source](https://docs.conare.ai/api-reference/connectors/connect-a-source.md): Connect a source whose catalog entry says `connect: "credentials"` (e.g. `twenty`): your app collects the end user's credentials and posts them here.
- [Create a connect link](https://docs.conare.ai/api-reference/connectors/create-a-connect-link.md): Mint a single-use hosted consent URL for a source. Send your user there; after consent the browser returns to your `redirectUrl`.
- [Disconnect a source](https://docs.conare.ai/api-reference/connectors/disconnect-a-source.md): Disconnect a source: revoke its OAuth grant and delete the memories it indexed for this end user.
- [List connected sources](https://docs.conare.ai/api-reference/connectors/list-connected-sources.md): List the end user's connected data sources and their sync status.
- [List syncable collections](https://docs.conare.ai/api-reference/connectors/list-syncable-collections.md): List the collections a connected source exposes (a CRM's `contacts`, `deals`, …). Use the ids to scope `collections` on link or sync.
- [List the source catalog](https://docs.conare.ai/api-reference/connectors/list-the-source-catalog.md): Every source type an end user can connect (~200): id, display name, and logo — everything a source picker needs.
- [Trigger a sync](https://docs.conare.ai/api-reference/connectors/trigger-a-sync.md): Start a background sync of a connected source into the end user's memory.
- [Create embeddings](https://docs.conare.ai/api-reference/embeddings/create-embeddings.md): Embed up to 96 texts with the same self-hosted model that indexes memories, so your own vector features live in the exact representation Conare retrieves with.
- [Check status](https://docs.conare.ai/api-reference/integration/check-status.md): A side-effect-free health check for your integration. Call it from deploy pipelines: it validates the key, namespace routing, and the memory-plane hop without creating an end user or consuming any quota.
- [Batch upsert by external ID](https://docs.conare.ai/api-reference/memories/batch-upsert-by-external-id.md): Apply up to 100 versioned source upserts in one call — the bulk path for mirroring a system of record.
- [Delete a container](https://docs.conare.ai/api-reference/memories/delete-a-container.md): Delete every memory carrying this container tag in the end user's space — retire one workspace's data (`workspace:acme`) without touching the rest.
- [Delete a memory](https://docs.conare.ai/api-reference/memories/delete-a-memory.md): Permanently delete one memory — document and vectors — by the `id` returned from save, search, and recall.
- [Delete a memory by external ID](https://docs.conare.ai/api-reference/memories/delete-a-memory-by-external-id.md): Delete a source-owned memory with a durable, versioned tombstone.
- [Get a memory by external ID](https://docs.conare.ai/api-reference/memories/get-a-memory-by-external-id.md): Fetch a source-owned memory — active or tombstoned — together with its latest durable receipt. Useful for verifying convergence after a migration or a replayed batch.
- [Save a memory](https://docs.conare.ai/api-reference/memories/save-a-memory.md): Save one distilled observation — a sentence or a paragraph, not raw event logs — to the end user's memory.
- [Save memories in batch](https://docs.conare.ai/api-reference/memories/save-memories-in-batch.md): Import up to 100 memories for one end user in a single request — the bootstrap path for existing history.
- [Search memories](https://docs.conare.ai/api-reference/memories/search-memories.md): Vector + full-text search over one end user's memory, fused and reranked. No LLM pass — sub-second.
- [Upsert a memory by external ID](https://docs.conare.ai/api-reference/memories/upsert-a-memory-by-external-id.md): Create or update a memory that mirrors a record your system owns. Identity is `(endUserId, source, externalId)`; bump `version` whenever the source record changes.
- [Get suggestions](https://docs.conare.ai/api-reference/recall/get-suggestions.md): Given a moment of context ("user just opened their dashboard"), returns up to 5 concrete next actions, each grounded in a specific memory — never generic advice (~3–6 s: fetch async, don't block render).
- [Recall](https://docs.conare.ai/api-reference/recall/recall.md): Everything relevant that's known about this user, synthesized into a citation-grounded answer — built to inject into your agent's system prompt at session start (~3–6 s).
- [Delete a user](https://docs.conare.ai/api-reference/users/delete-a-user.md): Permanently erase an end user: their entire memory store, their connector grants (revoked so nothing can sync back in), and their dashboard roster entry.
- [Update a user](https://docs.conare.ai/api-reference/users/update-a-user.md): Label an end user with a display name and/or email so your partner dashboard shows something readable instead of an opaque hash.
- [Migrate from Turbopuffer](https://docs.conare.ai/db/migrate-turbopuffer.md): Copy every Turbopuffer row, vector, and attribute into ConareDB with a resumable, verifying importer that stays read-only on the source.
- [Namespaces & keys](https://docs.conare.ai/db/namespaces.md): Namespace isolation, lifecycle states, scoped cdb_ API keys with read, write, and delete permissions, per-key rate limits, and metadata endpoints.
- [ConareDB](https://docs.conare.ai/db/overview.md): ConareDB is an object-storage-native database with hybrid vector, BM25 keyword, property, and graph search across isolated namespaces.
- [ConareDB quickstart](https://docs.conare.ai/db/quickstart.md): Provision a ConareDB key, upsert vectored rows into a namespace, and run a hybrid vector plus keyword plus chronological search in three requests.
- [Search](https://docs.conare.ai/db/search.md): Run vector ANN, BM25 keyword, and chronological ranked branches over one filtered row set with a typed property filter DSL and per-branch top_k.
- [Writes & bulk ingest](https://docs.conare.ai/db/writes.md): Row upsert, prop patch, and delete semantics, JSON and base64 fp16 vector encodings, background compaction, and the binary bulk-vectors ingest path.
- [Embeddings API](https://docs.conare.ai/embeddings/api.md): Stateless POST /api/v1/embeddings endpoint returning unit-normalized document vectors from the same model Conare retrieves with, with model pinning.
- [Custom models](https://docs.conare.ai/embeddings/custom-models.md): An embedding model fine-tuned on your data, acceptance-gated against a frozen evaluation set of real queries, and promoted only on a measured win.
- [Embeddings](https://docs.conare.ai/embeddings/overview.md): Retrieval embeddings from the same model Conare indexes with — served as a hosted API today, or fine-tuned on your corpus and evaluation-gated.
- [Conare Platform](https://docs.conare.ai/index.md): Conare gives your app custom retrieval infrastructure: per-user AI memory, a vector plus full-text database, hosted embeddings, and data connectors.
- [Integrations](https://docs.conare.ai/integrations/overview.md): Six first-class connectors — Notion, Granola, GitHub, Slack, Gmail, Twenty CRM — plus a live catalog of 200 more sources end users can connect.
- [Source-owned memories](https://docs.conare.ai/memory/lifecycle.md): Upsert and delete source-of-record memories by (endUserId, source, externalId, version) with idempotent, crash-safe, out-of-order-safe delivery.
- [MCP tools](https://docs.conare.ai/memory/mcp.md): Connect Claude Code, Codex, Cursor, and other MCP clients to Conare memory for recall, search, save, and forget tools over Streamable HTTP.
- [AI Memory](https://docs.conare.ai/memory/overview.md): Per-end-user persistent memory with save, hybrid vector plus BM25 search, LLM-synthesized deep recall, proactive suggestions, and connector ingest.
- [Authentication](https://docs.conare.ai/platform/authentication.md): Authenticate partner API requests with cint_ Integration Bearer keys — scopes, one-time secrets, base URLs, and separation from ConareDB cdb_ keys.
- [Core concepts](https://docs.conare.ai/platform/concepts.md): Integrations, end-user IDs, per-source containers, opaque tenancy, and the three retrieval modes — search, deep recall, and suggestions.
- [Errors & limits](https://docs.conare.ai/platform/errors.md): The stable error envelope, X-Request-Id correlation, HTTP status code meanings, rate-limit headers, and 402 quota-exhaustion behavior.
- [AI Memory quickstart](https://docs.conare.ai/platform/quickstart.md): Create an Integration key, verify it with a status check, save an end user's first memory, and recall it via search or synthesized deep recall.
- [TypeScript SDK](https://docs.conare.ai/platform/sdk.md): Typed, zero-dependency TypeScript SDK for the Conare partner API — works in Node, Bun, Deno, and Cloudflare Workers on the global fetch runtime.

## OpenAPI Specs

- [openapi](https://docs.conare.ai/api/openapi.yaml)