Install with your AI

Let your agent install Kvendra

Kvendra ships a machine-readable install guide following the llms.txt convention. It contains the whole setup — Platform up, embeddings, MCP wiring, Skills, and the optional CLI — condensed into unambiguous, copy-paste steps, in order, with a verification after each one. An agent can read it and run the setup end-to-end.

Point your Claude Code at the guide and ask:

Read https://kvendra.dev/llms.txt and install and configure Kvendra for me.

You can also read it yourself: kvendra.dev/llms.txt.

What the agent will do

The guide is platform-first, so the agent starts with the one required piece — the self-hosted knowledge base — and layers the rest on top:

  1. Check prerequisites. Docker with the daemon running, Docker Compose v2, git and curl — and it tells you exactly what to install if one is missing.
  2. Bring up the Platform. Reuse a reference-stack clone you already have, or clone it, start it with up.sh, and confirm the health check on port 7777.
  3. Choose embeddings. Ollama as a local container (default, nothing leaves your machine), your own OpenAI-compatible embeddings provider, or a mock provider for CI.
  4. Connect Claude Code over MCP. Read the token, register the Platform with claude mcp add --transport http, ask you to restart Claude Code once, and prove the connection with a real read — which gives the agent the 14 knowledge-base tools.
  5. Install the Skills and onboard. Add the Claude Code plugin, onboard your project with /kvendra-skills:onboard-project, and from then on work with /kvendra-skills:kvendra.
  6. Optionally add the CLI broker. A zero-knowledge vault and capability broker for brokered operations — optional and in alpha.

Honest about where this is

The Platform and Skills are the parts you run today. The CLI broker is optional and in alpha — Skills and the Platform work without it, and the guide marks it clearly so your agent does not treat it as required. Everything the guide references is open-core: the Platform is AGPL, the CLI is Apache-2.0, the Skills are MIT, and the reference stack is MIT.

Prefer to follow along by hand? The same steps live in the Quickstart.