LabscoMCP early access ↗
george-bobbyMCP SERVER

Modal MCP

Deploy, run and debug Modal apps from your assistant, using the Modal CLI credentials already on your machine.

by george-bobby ↗PaaS Deployment & App HostingCommunity

LABSCO SUMMARY

Log search with context is the tool you'll actually reach for.

Deploy and run are the headline, but the one that changes a debugging session is search_modal_logs: it pulls the logs once and greps them locally, so you get surrounding lines, regex and match counts rather than the bare matching line the CLI's own search returns. The repository is candid that a few tools are escalation paths — deploy_modal_app, put_modal_volume_file, exec_modal_container — and says to keep them behind your client's approval prompts.

What it is

A server that drives the Modal platform — apps, containers, volumes and secrets — by shelling out to your local modal CLI. Because it uses whatever profile is already configured in ~/.modal.toml, there is no second set of credentials to issue or rotate.

What you get

  • 26 tools grouped by area, most of them taking an optional env argument to target a specific Modal environment
  • deploy_modal_app deploys an app and returns live, shareable endpoint urls; run_modal_app runs one function or entrypoint and streams the output
  • get_modal_app_logs and get_modal_container_logs fetch or follow logs with since, until, tail, search, source and timestamps
  • search_modal_logs greps an app's or container's logs and returns each hit with context_lines around it (default 3), plus a match_count — regex, case control and an exclude filter included
  • Containers: list_modal_containers, exec_modal_container, stop_modal_container
  • Volumes: list_modal_volumes, list_modal_volume_contents, copy_modal_volume_files, put_modal_volume_file, get_modal_volume_file, remove_modal_volume_file, plus create_modal_volume, rename_modal_volume, delete_modal_volume
  • Secrets and discovery: list_modal_secrets, create_modal_secret (values redacted from the echoed command), delete_modal_secret, get_modal_profile, list_modal_environments

Requirements

Python 3.11 or higher, uv for uvx, and the Modal CLI 1.x already authenticated — run modal setup if not. Run it as uvx mcp-modal (PyPI package mcp-modal, version 0.2.4, stdio transport). Deploy and run support additionally requires the target project to use uv with modal installed in its virtualenv. Optionally set MCP_MODAL_ALLOWED_LOCAL_PATHS to a list of directories; the two volume tools that touch local files then refuse any path outside them.

Setup effort

One command plus credentials — uvx mcp-modal, then supply credentials

See installation options