Marshal / by Local Logic Management
DevOps Automation

Commit to
any platform.

Marshal is an MCP server that gives any AI assistant real hands on your release process — stage and push code, run your test suite, and deploy to whatever's configured, without the assistant ever needing platform-specific instructions twice.

View on GitHub → See it work
Live Demo

One instruction,
three stages.

A single request to an assistant connected to Marshal — test, commit, push, deploy — with nothing platform-specific for the assistant to know in advance.

assistant · connected to marshal
> "Run the tests for kronos, and if they pass, push and deploy it."
→ run_tests { project: "kronos" } ✓ 42 passed, 0 failed (3.1s)
→ git_commit_push { project: "kronos", message: "fix: correct risk calc rounding" } ✓ pushed 1 commit to main
→ deploy { project: "kronos" } ✓ live at kronos.locallogicmanagement.com
How It Works

One protocol.
Every target.

Every project is a config entry, not a rewrite. Marshal reads it, picks the right adapter, and talks to whatever's on the other end.

AI ASSISTANT Claude, etc. MCP SERVER Marshal GIT HOSTS GitHub · GitLab TEST RUNNERS npm · pytest · go test DEPLOY TARGETS Vercel · Netlify · Docker · SSH
Capabilities

What Marshal
hands your assistant.

Four tool groups, one config file per project.

Code

Git & Code Push

Stage, commit, branch, and push. Opens GitHub PRs or GitLab MRs — via the CLI you already use, or a REST fallback.

git_status · git_commit_push · git_open_pr
Quality

Testing

Runs whatever your project already uses — npm test, pytest, go test — and reports pass/fail with the real output.

run_tests · exit codes · full logs
Release

Deployment

One tool, five providers: Vercel, Netlify, Docker, SSH, or a custom script — chosen per project by config, not by code.

deploy · pluggable providers
Platform

Config-Driven

Adding a new platform is a config entry and a small adapter function. Every project stays independent of every other.

devops-mcp.config.json · zero rewrites