kestra-io/mcp-server-python

Why STEADY

STEADY (74) because this is a coherent, well-documented MCP server from a credible open-source orchestration vendor, with a clear agent-consumption path (container image + MCP client configs for Claude/Cursor/VS Code + OSS and Enterprise auth modes). It is not rated higher for two reasons: Hlido's capture resolved to the Kestra AI Agents blog rather than the MCP server's repository, so license, releases, CI, maintenance recency and star count for this specific repo were not independently verified; and this is a public/repo-surface review with no hands-on behavioral run against a live Kestra instance (medium confidence). A confirmed repo-health pass plus a behavioral run would likely move it into the upper STEADY range.

Public-surface checklist

• PASS Repo reachable (required) — kestra-io/mcp-server-python referenced as the canonical product surface; README content captured via scout-enrich preanalysis.
• PASS Readme present (required) — README documents Docker distribution and MCP client configuration (OSS and EE); captured markdown preview well over 500 chars.
• PASS Install documented (required) — Docker run command plus MCP settings JSON for Cursor/Claude/VS Code documented for both OSS (username/password) and EE (API token) users.
• PASS Agent consumable — Model Context Protocol server; ghcr.io/kestra-io/mcp-server-python image; mcpServers config blocks for Cursor, Claude and VS Code.
• PASS Active 12mo (required) — Kestra AI Agents (which lists MCP clients as a current tool) last updated Sept 2025; © 2026 Kestra Technologies. NOTE: repo-specific push date was not captured; platform-level activity is the basis here.
• FAIL License present (required) — License file not visible in captured evidence — the capture landed on the Kestra AI Agents blog, not the repo page. Not independently verified for this repo.
• FAIL Releases present — Not captured; the container image is pinned to :latest in the README, but tagged repo releases were not verified.
• FAIL Community traction — Star count for this repo not captured (preanalysis gh_stars is null). The 27,793 stars on the blog belong to the main kestra-io/kestra platform repo, not the MCP server, and are not credited here.
• FAIL Ci or tests — Not captured; repo page was not crawled in this run.
• FAIL Recent commit 90d — Repo push date not captured; not independently verified for this repository.

What we saw

1 screenshot captured by the Hlido engine during the reviewed run (run-4fb8482b0478e9bc-kestra-io). Our own captures — not vendor marketing material.

×

←→

What it does well

• Genuinely agent-first: a Model Context Protocol server that exposes Kestra orchestration directly to MCP clients (Claude, Cursor, Windsurf, VS Code)
• Container-first distribution with a pinned image (ghcr.io/kestra-io/mcp-server-python) that avoids local Python/dependency management
• Copy-paste MCP client configuration is documented for both open-source and Enterprise Edition users
• Clean auth separation — OSS username/password vs Enterprise API-token config — with env-var controls for disabling tools and log level
• Backed by Kestra, an established open-source declarative-orchestration platform with Enterprise and Cloud editions

What it fails at

• Only useful to teams already running Kestra — it is a bridge to a backend, not a standalone tool
• The documented OSS config example embeds plaintext admin username/password; deployers must harden this with the token path
• Repo-health signals (license file, tagged releases, CI, push recency, star count) were not captured for this specific repository
• No live/browsable demo of the MCP tools; behaviour is asserted by the README, not independently exercised here
• Requires Docker plus a reachable Kestra instance (KESTRA_BASE_URL), so setup is heavier than a hosted agent tool

Related agents

• Activepieces Workflow & Automation · 90
• Dust Workflow & Automation · 90
• Integrately Workflow & Automation · 90

Agent relevance

API MCP

An MCP client (Claude Desktop, Cursor, Windsurf, VS Code) runs the container image ghcr.io/kestra-io/mcp-server-python via a stdio Docker command, pointed at a Kestra instance (KESTRA_BASE_URL / tenant) and authenticated with either OSS username-password or an Enterprise API token. The agent then invokes Kestra orchestration through the exposed MCP tools; specific tools can be disabled via KESTRA_MCP_DISABLED_TOOLS. Requires a reachable Kestra backend and Docker.

Agent-friendly score: 8/10

Evidence

• You can run the MCP Server in a Docker container (image ghcr.io/kestra-io/mcp-server-python:latest) — source (2026-09-27) verified
• MCP client configuration is documented for Cursor, Claude and VS Code ("Paste the following configuration into your MCP settings") — source (2026-09-27) verified
• Separate configurations for open-source users (username/password) and Enterprise Edition users (KESTRA_API_TOKEN), with env-var controls for disabled tools and log level — source (2026-09-27) verified
• Backed by Kestra, an established open-source declarative orchestration platform with Enterprise and Cloud editions — source (2026-09-27) verified
• Repository health signals for this specific MCP server repo (license file, tagged releases, CI, push recency, star/fork counts) — source (2026-09-27)