kvendra.dev · open core

Stop your AI agent from re-deciding your project.

Your AI coding agent loses the thread of your project and re-decides things you already settled. Kvendra gives it a structured, queryable knowledge base of your project — plus traceable pipelines — so it stops re-deciding and you stay in control.

Ask “how do we avoid leaking tokens” and it surfaces the relevant decision even if those exact words never appear — because Kvendra finds the right thing by meaning, not by keyword.

Get Started Install with your AI View on GitHub

Your first 10 minutes — all inside Claude Code, no terminal: 

# in Claude Code
/plugin marketplace add KvendraAI/kvendra-skills
/plugin install kvendra-skills@kvendra-marketplace
/setup self-hosted

You need Docker running. The wizard does the rest: it finds an existing reference-stack clone or clones one for you, brings it up with local embeddings, and wires Claude Code.

See what you actually get out of it →

How it's put together

Three open-core pieces: Platform is the required core — the knowledge base. Skills are your day-to-day — pipelines run as one-command workflows on top of it. The CLI is a foundational security layer that is optional to get started; add it when you want your agent to run real ops (a git push, a deploy) without seeing your tokens.

KB Engine (Kvendra Platform)

A knowledge base your agent can actually query.

  • PostgreSQL + pgvector semantic search
  • Typed entities with stable ids
  • Transactions and a server-side changelog
  • Self-hostable — the core runs on your own infra

Skills for AI agents

Turn pipelines into one-command workflows.

  • A Claude Code plugin that runs on the Platform
  • Thin skills — the knowledge lives in the KB
  • Read at runtime, portable across projects

CLI & Broker — optional security layer

Add a zero-knowledge vault and a capability broker.

  • Optional: Skills and Platform work without it
  • A zero-knowledge vault for your secrets
  • A capability broker that runs ops on your behalf
  • Pre-alpha — add it when you need brokered ops

Why Kvendra

Open-core

The CLI, the KB engine and the skills are open source — read them, fork them, run them.

Self-hostable

Bring up your own single-tenant knowledge base; nothing has to leave your infrastructure.

Your Postgres/pgvector

Knowledge lives in a database you control — relational data and the vector index in one store.

MCP-native

Tools and the broker speak MCP, so your agent talks to Kvendra through an open protocol.