Quill

Quill is RavenDB's context layer for putting production AI agents on top of existing PostgreSQL, SQL Server or MySQL systems without migrating them. It connects via change data capture, generates embeddings, vector indexes and full-text search automatically, and adds governed per-agent permissions, persistent memory and multi-channel delivery, so teams ship agents in weeks rather than assembling a RAG stack.

Quill is RavenDB's context layer for putting production AI agents on top of SQL databases a company already runs, without migrating the system of record. It connects to PostgreSQL, Microsoft SQL Server or MySQL through change data capture — passive log reads rather than table scans or heavy queries against the live system — and maintains a continuously synchronised, AI-ready copy alongside it. From that copy Quill automatically generates embeddings, vector indexes and full-text search, so there is no separate vector database to run, and layers persistent cross-session memory, retrieval-augmented generation and the agents themselves on top. Access is governed per agent: each agent is synced only the slice of data it is authorised to see, with permissions defined independently of the underlying database's access controls, which is the part most hand-built RAG deployments get wrong and only discover after an agent answers a question it should not have been able to answer. Quill is model-agnostic — bring your own LLM provider and switch later — and surfaces agents through web chat, WhatsApp, Telegram, Slack and Discord. It runs in RavenDB's cloud, self-hosted, or fully on premises including air-gapped environments, so data residency and compliance control stay with the customer. RavenDB launched Quill on 8 September 2026, pitching roughly six weeks to production against the 18 to 24 months a comparable in-house build typically takes across security review, infrastructure and re-indexing. The vendor is not a startup: RavenDB was founded in 2010 by CEO Oren Eini and CTO Paweł Pekról, ships a NoSQL document database it reports is used by more than 12,000 customers across 50 industries, and now sells Quill alongside its on-premise, cloud and edge database products. Pricing is published and metered on writes rather than seats, with a three-month full-feature free trial.

Category

Enterprise Search & Knowledge

Pricing

Subscription, Usage-based

Target Market

CTOs, VPs of Engineering, Enterprise Architects, Data Engineers, CIOs

Deployment

Cloud-first, Self-hosted, Hybrid

Founded

2010

Headquarters

Hadera, Israel

Customers

RavenDB reports 12,000+ customers across 50 industries for its database products; Quill's own customer count is not public, with Albos Technologies and Holdings named as an early adopter

• ✓
  CDC connection to existing SQL

  Connects to PostgreSQL, SQL Server or MySQL via change data capture with passive log reads, so the production database sees minimal load.

• ✓
  Automatic AI pipeline

  Generates embeddings, vector indexes and full-text search from the synced data, removing the need to run a separate vector database.

• ✓
  Governed per-agent exposure

  Each agent is synced only its authorised slice of data, with permissions set independently of the source database's access controls.

• ✓
  Persistent memory

  Context carries across sessions, with long-term memory, retention policies and conversation forking and rewinding on higher tiers.

• ✓
  Bring your own model

  Works with any LLM provider and lets teams switch models later without rebuilding the retrieval and governance layer.

• ✓
  Multi-channel delivery

  Agents are exposed through web chat, WhatsApp, Telegram, Slack, Discord and voice without building a separate front end per channel.

• ✓
  On-premises and air-gapped deployment

  Runs fully in the customer's environment so regulated buyers keep data residency and compliance control in house.

• •
  Customer support agent over legacy order data

  Expose order, billing and account tables to an agent that answers customer questions without migrating the transactional system.

• •
  Internal knowledge agent across several databases

  Unify multiple SQL sources into one context layer so one agent can answer questions spanning systems that never talked to each other.

• •
  Regulated on-premises deployment

  Run retrieval, memory and agents inside an air-gapped environment where sending enterprise data to a hosted AI vendor is prohibited.

• •
  Replacing an in-house RAG build

  Swap a hand-assembled CDC, embedding, vector-store and permissions stack for one product, cutting time to production from many months to weeks.

• •
  Messaging-channel agents

  Deliver the same governed agent through WhatsApp, Slack or Telegram for field staff and customers without separate integrations.

Metered in write request units rather than seats: every document write counts as one WRU — ingesting a record, updating data, or an agent sending a message — while reads and queries are always free and document size and ingestion speed do not affect the bill. Starter is $499/month for 1M WRU with $0.60 per extra 1,000; Pro is $1,799/month for 8M WRU with $0.30 overage; Enterprise is quoted. LLM token costs are billed separately by whichever model provider you bring, and the three-month trial includes all Pro features with unlimited WRU.

This page was written from 5 sources, 2 on domains other than ravendb.net.

1. 1.ravendb.net — quillvendor
2. 2.ravendb.net — pricingvendor
3. 3.ravendb.net — aboutvendor
4. 4.techstartups.com — ravendb launches quill to bring production ai agents to ente
5. 5.hn.algolia.com — search