kbagent for AI agents

kbagent is built to be driven by AI coding agents (Claude Code, Cursor, Copilot), not just humans. It gives an agent a stable command surface, a machine-readable reference, and safety rails so it can operate Keboola without you handing over unrestricted access.

Claude Code plugin

Add kbagent to Claude Code

Run these two commands inside Claude Code to install the plugin from Keboola’s AI Kit marketplace (use the copy button on the block):

/plugin marketplace add keboola/ai-kit
/plugin install kbagent@keboola-claude-kit

The plugin installs via Claude Code’s marketplace, not as a downloaded file — so it’s a command you run in the assistant, not a button. It adds a /keboola slash command that spawns a keboola-expert subagent with fresh context and hard rules (fetch the current reference, dry-run first, prefer the CLI over raw REST/MCP, gate on version), plus a structured verification payload. kbagent doctor tells you whether the plugin is installed.

The plugin also adds /kbagent:setup, which does the whole first run in one go: it installs the CLI if it is missing, checks whether a project is already connected and skips ahead if so, otherwise signs you in through the browser and registers every project the session can reach, then runs kbagent doctor and suggests what to try first. It is safe to run twice. Pass a stack URL or a project name as its argument to aim it.

/kbagent:setup is a plugin slash command, which has two consequences worth knowing before you follow the steps below. It exists only after the plugin is installed, so the two /plugin lines are never optional. And it runs only inside Claude Code or Claude Desktop chat, never in a shell.

Set up your client

Claude Code

1. Start Claude Codeclaude
2. Add the marketplace/plugin marketplace add keboola/ai-kit
3. Install the plugin/plugin install kbagent@keboola-claude-kit
4. Run the setup command with your stack URL/kbagent:setup https://connection.keboola.com

Four steps, and step 4 covers the rest: it installs the CLI, signs you in through the browser, registers your projects and verifies the result. There is no token to generate or paste.

Claude Desktop

Same shape, but the marketplace and plugin steps go through Claude Desktop’s own plugin UI rather than slash commands.

1. Open Settings → Plugins → Add → Add marketplace, paste keboola/ai-kit, and click Sync.
2. In the plugin list, enable the plugin named exactly kbagent. Don’t pick keboola-cli in that same list: that is the unrelated legacy kbc CLI.
3. In the chat, run /kbagent:setup https://connection.keboola.com.

Cursor, VS Code, Codex, Devin Desktop

No kbagent plugin exists for these clients yet, so /kbagent:setup is unavailable to them: it is a Claude Code plugin command. Set up by hand instead.

1. Open the client.
2. Install the CLI, following the tab for your operating system in Get started.
3. Run kbagent context so the agent learns the command surface.
4. Connect your project.
5. Verify with kbagent doctor.

Plain terminal

No AI client involved: install the CLI, connect your project, and verify with kbagent doctor.

The context reference

Any agent — plugin or not — should start by loading the full command reference:

kbagent context

It prints usage instructions and the complete, version-matched command list as Markdown, so the agent works from what your installed version actually supports rather than guessing.

Sandbox the agent

Don’t give an agent write access it doesn’t need. kbagent’s permission firewall lets you scope a session or a workspace:

# Read-only local workspace for an agent
kbagent init --from-global --read-only

# Or block writes for a single session
kbagent --deny-writes config list

Combined with per-project scoped tokens (kbagent token …), you can let an agent explore and propose changes while you keep approval of anything destructive.

Set a conversation ID

Set a conversation ID before the agent runs commands, so platform observability can correlate the whole session (every request carries an X-Conversation-ID header):

export KBAGENT_CONVERSATION_ID="<unique-id>"

How it fits with the other AI tools

• kbagent — the agent’s hands on your projects from the terminal, with sandboxing.
• MCP server — direct tool calls over MCP; kbagent can also call MCP tools via kbagent tool.
• AI Kit — coding-assistant plugins for building Keboola components and apps.
• Kai — the in-product assistant; kbagent kai ask -m "why did last night's load fail?" puts the same assistant in your shell (beta).

---

Next: Command reference →