Help us improve
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
Share bugs, ideas, or general feedback.
By keboola
Manage Keboola Connection projects through an AI-friendly CLI interface, automating configs, jobs, lineage, branches, GitOps sync, and PR reviews with enforced best practices like fresh fetch, dry-run, and CLI-over-REST.
npx claudepluginhub keboola/cli --plugin kbagentDelegate a Keboola Connection task to the keboola-expert specialist subagent (enforces fresh-fetch, dry-run, CLI-over-REST, version gate).
Run the kbagent-pr-reviewer subagent against the PR for the current git branch (or a specified PR number/URL). Posts a single structured comment review via `gh pr review --comment --body-file`. Read-only on the working tree; never approves or requests changes on GitHub.
Read-only code reviewer for kbagent (keboola-agent-cli) pull requests. MUST BE USED when the user runs `/kbagent:review`, asks "review PR
Keboola Connection ops specialist. Enforces fresh-fetch, dry-run, CLI-over-REST, version gate and confirmed apply.
Uses power tools
Uses Bash, Write, or Edit tools
Share bugs, ideas, or general feedback.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge.
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Keboola project management and review toolkit with 10-agent review team, CLI sync commands, and financial intelligence analysis
Data engineering plugin - warehouse exploration, pipeline authoring, Airflow integration
The most comprehensive SAP Datasphere plugin for Claude. 18 specialized skills covering exploration, data modeling, integration, BW Bridge migration, security architecture, CLI automation, business content activation, catalog governance, performance optimization, and troubleshooting — all through natural language. Powered by 45 MCP tools with enterprise-grade security.
Claude Code skill pack for ClickHouse (24 skills)
Skills for drt — create syncs, debug failures, initialize projects, and migrate from Census/Hightouch
This plugin provides a specialized suite of skills for data engineers and database practitioners working on Google Cloud. It acts as an expert assistant, allowing you to use natural language prompts in your preferred coding agent to architect complex data pipelines, transform data with dbt, write Spark and BigQuery SQL notebooks, and orchestrate end-to-end workflows across GCP's data ecosystem.
Keboola project management and review toolkit with 10-agent review team, CLI sync commands, and financial intelligence analysis
Toolkit for building and deploying data apps to Keboola — Streamlit development with validate/build/verify workflow, plus deployment guides for Node.js, Python, and any web framework
Comprehensive toolkit for building Keboola Python components with Agent Skills format. Includes specialized skills for component development, UI/schema design, unified testing (datadir, unit, VCR), debugging, consolidated code quality and backward compatibility review, and platform context knowledge.
Group demo customers into cohorts and track repeat purchases.
Answer revenue questions over the demo orders dataset — no SQL required.
One CLI to manage all your Keboola projects. Designed to be driven by AI agents -- Claude Code, Codex, Gemini, Cursor -- but works great standalone too.
No more switching between the UI, old CLI, MCP server, and raw API calls. kbagent wraps everything into workflow-oriented commands where dev branches propagate automatically, multi-project operations run in parallel, and AI agents can be sandboxed safely.

uv tool install git+https://github.com/keboola/cli
Auto-updates kbagent and its keboola-mcp-server dependency on every launch (since 0.30.1) -- no more silently running on a six-month-old MCP server. Run kbagent changelog to see what changed.
Want a browser dashboard? One command:
uv tool install --with 'keboola-agent-cli[server]' 'git+https://github.com/keboola/cli'
kbagent serve --ui
# Open the URL printed at startup -- the browser is auto-authenticated.
The React SPA is bundled inside the wheel by a hatchling build hook (requires Node 20+ on the install host so npm run build can run during wheel creation). Single Python process at runtime; no Node needed once installed. Covers everything the CLI exposes (projects, configs, storage, jobs, flows, schedules, MCP tools, lineage, scheduled AI agents with cost/token timeline). Agent runs that produce long-form reports (e.g. "Storage Cleanup Advisor", "Schedule Drift Detector") surface in a dedicated Artifacts tab — GFM-rendered preview in a VSCode-style viewer with one-click Copy / Download .md for hand-off to Slack, Notion, or your editor. See web/README.md for the dev-mode setup with hot reload.
Schedule AI agents to run inside kbagent serve -- cron, manual triggers, or chained (one agent finishes, another starts). Each task picks one of three action flavours:
claude / codex / gemini with a custom prompt. The subprocess inherits KBAGENT_SERVE_URL + KBAGENT_SERVE_TOKEN so it calls back via kbagent http get /... instead of forking fresh CLI processes against stale config.keboola-mcp-server, validated against its input schema.kbagent ... command with its args (encrypted secrets supported).Every run is recorded as a persisted timeline (JSONL on disk, 0600) with authoritative cost & token accounting (Opus 4.7 / Sonnet 4.6 / Haiku 4.5 pricing built-in) and per-step replay over SSE. Long-form markdown reports (e.g. "Storage Cleanup Advisor", "Schedule Drift Detector") auto-surface in a dedicated Artifacts tab with GFM preview + one-click Copy / Download .md for Slack, Notion, or your editor.
Build the agent once, schedule it, walk away — the platform handles auth, scheduling, history, cost reporting, and report rendering.
The scheduler runs inside kbagent serve -- the same single Python process that hosts the Web UI. If you already installed kbagent with the [server] extras (see Web UI above), you're set; otherwise:
uv tool install --with 'keboola-agent-cli[server]' 'git+https://github.com/keboola/cli'
kbagent serve --ui
# Open the URL printed at startup -> sidebar "Agent Tasks" -> "+ New task".
Keep kbagent serve running for the scheduler to fire CRON triggers. Run history is persisted (JSONL on disk), so closing the server doesn't lose past runs -- it just pauses future scheduling until you restart. Architecture detail and the full endpoint reference live in docs/web-server.md.
This CLI is built AI-first. Every command outputs structured JSON (--json), errors include machine-readable codes, and the permission firewall enforces safety at the code level -- not via prompt instructions.
Claude Code plugin (agent learns all 100+ commands + gets a specialist subagent for writes):
/plugin marketplace add keboola/cli
/plugin install kbagent@keboola-agent-cli
Then either let the kbagent skill auto-trigger from natural prompts, or delegate explicitly with /keboola <task> -- the slash command spawns a kbagent:keboola-expert subagent with fresh context, hard rules (fresh fetch, dry-run first, prefer CLI over REST/MCP, version gate), and a JSON verification payload. See docs/TUTORIAL.md §6.
Any other agent -- just tell it to run kbagent context and it gets the full command reference.
What you can ask your agent:
"Give me a full inventory of all Keboola projects -- configs, jobs, components, data volumes."
"Find the last failed job in project X, figure out why it crashed, spin up a workspace with the input data, and fix the SQL."
"Compare the SQL transformation between production and the dev branch."