Connect
Write documents, upload files or connect an app. Each lands in a namespace.
Context infrastructure
Everything your agents need to know.
Conare searches all of your company's data and puts the right answer first. Built for accuracy, and measured on public benchmarks.
Context
- Documents
- Files
- Apps
How it works
Connect. Sync. Retrieve. Deliver.
Sync
A changed record replaces its old version. A deleted one is deleted.
Retrieve
Keyword and vector search find candidates. The reranker puts the best first.
Deliver
Your model gets the passages that answer the question, with their source.
Quickstart
Add a document. Query it.
Conare chunks, embeds and indexes on write, keeps your data updated in real time, and reranks every query.
$ npm install conare
import { Conare } from "conare";
const docs = new Conare().namespace("docs");
await docs.add({ text: "EU checkout broke after a key rotation." });
const [top] = await docs.query("why did checkout break?");
console.log(top.score, top.text);Output
0.91 EU checkout broke after a key rotation.Products
Everything retrieval needs, behind one key.
Use the platform end to end, or call the models on their own.
namespaces
Search
Write documents or upload files into a namespace. Each query runs keyword and vector search, filters on metadata and reranks the results.
Docs →sources
Sync
Connect HubSpot, Gmail, Slack and more. New and changed records are written, and deleted ones are deleted.
Docs →POST /parse
Parse
PDF, Word, PowerPoint, Excel, CSV and images, with OCR for scanned pages. Nothing is stored.
Docs →conare-embed-1
Embed
Separate models for documents and queries, through an OpenAI-compatible endpoint.
Docs →conare-rerank-1
Rerank
Scores each passage against the query, best first. On in every query, or on its own through a Cohere-compatible endpoint.
Evals →Docs →Sources
Connect the apps your team already uses.
Each source keeps one app in sync with a namespace, on its own schedule.
Sources
- Gmail
- Slack
- Notion
- Linear
- GitHub
- Granola
- Twenty
- GCS
- HubSpot
- Salesforce
- Attio
- Affinity
- Google Sheets
- and more
Operations
Nothing to run.
Conare runs the database, the models and the sync. You don't patch servers, rebuild indexes or plan capacity before a launch.
What you don't manage
- Servers, shards and index tuning
- Embedding and reranking models
- Sync jobs for every app
- Separating each organization's data
- Deleting a user: one call removes everything Conare holds of them
Benchmarks
Accuracy you can check.
conare-embed-1 and conare-rerank-1 on public benchmarks against the best models on each, Claude Opus 5.5 and GPT-6 Sol included.
ViDoRe v3, embedding
67.0
nDCG@10, higher is better · #1 of 9
- 1conare-embed-167.0
- 2Perplexity pplx-embed-v2-late-9b, page images66.5
- 3Perplexity pplx-embed-v2-late-9b, page text66.1
- 4Mixedbread search, page images64.2
- 5Mixedbread search, page text62.3
See the benchmark
ViDoRe v3, reranking
70.2
nDCG@10, higher is better · #1 of 10
- 1conare-rerank-170.2
- 2Voyage rerank-2.568.5
- 3Voyage rerank-2.5-lite67.4
- 4zerank-267.0
- 5zerank-166.7
See the benchmark
Legal case search
1.3%
% ranked wrong, lower is better · #1 of 6
- 1conare-rerank-11.3%
- 2Claude Opus 5.52.5%
- 3GPT-6 Sol7.6%
- 4Mixedbread mxbai-rerank-v3.111.4%
- 5Contextual rerank-v253.2%
See the benchmark
Several conditions at once
2.5%
% ranked wrong, lower is better · #2 of 6, level with Claude Opus 5.5
- 1Claude Opus 5.51.1%
- 2conare-rerank-12.5%
- 3GPT-6 Sol3.8%
- 4Mixedbread mxbai-rerank-v3.116.7%
- 5Contextual rerank-v250.0%
See the benchmark
Read 14 October 2026. Comparator scores are what each vendor or leaderboard published, or our run of open weights where marked.
Most accurate, and the cheapest.
On ViDoRe v3, conare-embed-1 scores higher than every embedding API on this chart and costs less per token than each of them.
See the benchmark →Contact
See it on your own data.
Conare is invite-only. Book a demo and we'll run your questions on your data with you.