# Quivr

> Open-source engine that turns continuous content streams into durable ingestion, hybrid search, and topic alerts via an HTTP API.

Quivr, from The Vibe Company, is an open-source engine for content that keeps arriving: news, feeds, mailboxes and posts. It stores incoming documents, processes them in the background so they can be searched, and notifies an application when a new document matches a followed topic. The documentation states it is at the evaluation stage, with a `v0` API that may change, and advises against production use.

## What It Is

Quivr is a backend engine, not a finished product: it has no screens or user accounts of its own. Developers build the interface, such as a newsroom tool, media monitoring service or research assistant, and call Quivr over an HTTP API. Documents are collected from RSS and Atom feeds, Microsoft 365 mailboxes and X lists, pushed by partners, or sent by the application as text or files such as PDFs. Search returns ranked passages linked to a document and version.

## How Content Flows

Quivr stores every version of every document, so corrections create new immutable versions instead of overwriting history. Text is keyword-searchable once segmented, and embeddings for meaning search arrive afterwards. Search can be by keywords, by meaning, or hybrid. Saved alerts can be keyword queries or plain-language descriptions; each match records its reason, and a signed webhook delivers identifiers to the application's alert receiver.

## Plugins and Extensibility

Jobs that depend on content are handled by plugins, which are small HTTP web services in any language. They cover file formats, source collectors, text preparation for search, result ranking and alert rules. Only text preparation and ranking are required, and Quivr ships a plugin for each. Go and Python SDKs and a plugin contract test runner are provided.

## Running and Connecting

Quivr runs on servers the operator controls, using a Go API and worker with PostgreSQL, S3-compatible storage, Temporal and Weaviate. The local stack runs on Linux x86_64 and Apple Silicon Macs with Docker. AI agents can search and cite content through an MCP server, and a `quivr search` CLI is included. A demo web app shows a monitoring dashboard built on the API. There is no OCR, so scanned documents stay unsearchable.

## Features
- Durable, idempotent document ingestion
- Immutable document versions for corrections
- Keyword, semantic and hybrid search
- Keyword and plain-language alerts with match evidence
- Signed webhook delivery with retries
- RSS/Atom, Microsoft 365 mail and X list connectors
- Plugin system with Go and Python SDKs
- MCP server for AI agents
- Change feed via polling and SSE
- Prometheus metrics and OpenTelemetry export
- Demo web app

## Integrations
PostgreSQL, Temporal, Weaviate, S3-compatible storage, Docker Compose, MCP, Microsoft 365, X, RSS and Atom, OpenAI, Cohere, TypeSafe Jev, Kubernetes, KEDA, Railway

## Platforms
WINDOWS, MACOS, LINUX, WEB, API, DEVELOPER_SDK, CLI

## Pricing
Open Source

## Links
- Website: https://docs.quivr.thevibecompany.co
- Documentation: https://docs.quivr.thevibecompany.co
- Repository: https://github.com/The-Vibe-Company/quivr
- EveryDev.ai: https://www.everydev.ai/tools/quivr