Creativity · MCP — server

MCP Kestra Server

The MCP Kestra Server integrates Kestra, an open-source event-driven orchestration platform that competes with Airflow, Dagster, and Prefect. It exposes flows, executions, namespaces, and triggers as MCP tools — useful for teams that build polyglot data pipelines in YAML and want an LLM copilot for operations.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: Python / TypeScript
Transports: stdio, http

Capabilities

Tools: list_flows, trigger_execution, fetch_logs, restart_execution, list_namespaces
Resources: kestra://flow/{namespace}/{id}, kestra://execution/{id}
Auth: basic or API token via KESTRA_API_TOKEN

Install

pipx install mcp-server-kestra

Configuration

{
  "mcpServers": {
    "kestra": {
      "command": "mcp-server-kestra",
      "env": {
        "KESTRA_URL": "https://kestra.example.com",
        "KESTRA_USER": "${KESTRA_USER}",
        "KESTRA_PASSWORD": "${KESTRA_PASSWORD}"
      }
    }
  }
}

Frequently asked questions

What makes Kestra different from Airflow?

Kestra is YAML-first, event-driven, and UI-heavy. Pipelines are authored as declarative YAML rather than Python DAGs, which pairs naturally with LLMs.

Can the LLM author flows?

Yes — since flows are YAML, Claude can draft a flow and push it via the create_flow tool (gated by approval). Combined with validation, this is a popular 'AI edits the pipeline' workflow.

Does it support Apache Kafka triggers?

Kestra has first-class Kafka and webhook triggers. The MCP server can list them and toggle enabled/disabled state.

Sources

Kestra — open-source orchestration — accessed 2026-04-20
Model Context Protocol — accessed 2026-04-20
Kestra GitHub — accessed 2026-04-20