Skip to main content
Glama

deployplan

Coding agents can write the app. They still dump Vercel + Supabase on a human who has never used either. That person creates two logins, hands the agent tokens, then pastes failure emails back into the chat.

deployplan is the other loop: the agent inspects the repo, asks scale and preference questions, shows how the deploy will look, and applies on one account. Open source. Not a paid product. The human does not pick vendors. When something breaks, the agent reads logs here.

What the human does

  • Answer who will use it, whether it can sleep, login/files/jobs, domain

  • Open the URL

Related MCP server: golive-mcp

What the human does not do

  • Pay or pick a plan

  • Choose Vercel, Supabase, Railway, or Neon

  • Create those accounts

  • Forward provider email to the agent

Agent install

Cursor MCP (.cursor/mcp.json):

{
  "mcpServers": {
    "deployplan": {
      "command": "npx",
      "args": ["tsx", "packages/mcp/src/index.ts"],
      "cwd": "/absolute/path/to/deployplan"
    }
  }
}

Claude Code and Codex should read AGENTS.md and skills/deploy-plan/SKILL.md. Point their MCP at the same command.

CLI

npm install
npx tsx packages/cli/src/index.ts inspect --root ./fixtures/next-postgres
npx tsx packages/cli/src/index.ts interview
npx tsx packages/cli/src/index.ts required --root ./fixtures/next-postgres --answers '{"audience":"just_me","availability":"hobby_sleep","features":["login"],"domain":"later"}'
npx tsx packages/cli/src/index.ts propose --root ./fixtures/next-postgres --answers '{"audience":"just_me","availability":"hobby_sleep","features":["login"],"domain":"later"}'
npx tsx packages/cli/src/index.ts apply --account local --root ./fixtures/next-postgres --answers '{"audience":"just_me","availability":"hobby_sleep","features":["login"],"domain":"later"}'

apply writes compose and starts Docker when it is installed: Postgres, Redis, MinIO, NATS, Mailpit, nginx for a static index.html, or an app with a Dockerfile. Connection strings stay on this account. The agent reads logs here.

HTTP API for agents that do not speak MCP: npm run api then POST http://127.0.0.1:8786/v1/apply.

Repo

Path

Role

catalog/

services, interview, what is required vs optional

docs/SERVICES.md

required vs optional vs Docker-live

packages/core

inspect → interview → topology → quote

packages/platform

control plane + Docker runtime

packages/mcp

MCP tools agents call

packages/api

HTTP tools agents call

packages/cli

same engine for humans and tests

skills/deploy-plan

Cursor / Claude / Codex skill

The control plane runs the bundle with Docker on the machine. There is no checkout and no second vendor signup.

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    A
    maintenance
    Enables AI agents to discover, configure, and manage local development servers. Provides tools for app registration, port allocation, lifecycle control, and log access without manual config editing.
    311 npm
    MIT
  • A
    license
    Not graded
    quality
    A
    maintenance
    Enables AI agents to deploy code to any hosting provider by creating PRs, building, and verifying health checks, all from a single natural language command.
    1
    MIT
  • F
    license
    Not graded
    quality
    D
    maintenance
    Provisions Docker-based development environments on demand, allowing AI agents to create, manage, and inspect containerized dev environments without manual setup.
    -
  • A
    license
    A
    quality
    C
    maintenance
    Enables AI coding agents to instantly provision ephemeral developer infrastructure like Postgres, Redis, MongoDB, and container deployments over HTTPS, with no Docker or signup required for the free anonymous tier.
    31
    MIT