dploy

Building software is faster than deploying it. dploy closes that gap.

Point dploy at a GitHub repo (or a local project), and it provisions an ephemeral sandbox, figures out how to run it, and hands you a live URL — without you touching a config file.

What It Solves

Deployment usually requires someone to inspect the repo, pick a runtime, identify install and start commands, find the right port, and wire everything together manually. AI tools have cut the time it takes to build software, but that deployment overhead remains.

dploy absorbs that step. It uses deterministic heuristics for common stacks (Node, Python, Go) and falls back to an agent-driven analysis pass for anything messier. Multi-service repos, CLI tools, and non-standard layouts are handled, not rejected.

Stack

Layer	Tech
Frontend	React, Vite, TanStack Query, Tailwind, xterm.js
Backend	FastAPI, SQLAlchemy, Pydantic
Sandbox runtime	Modal
Agent runtime	OpenClaw
Models	Claude Haiku 4.5 (configurable)
Auth	GitHub OAuth
CLI	TypeScript, Ink

Install the CLI

The fastest way to get dploy on your machine:

npm install -g @ryantanen/dploy

…or, without npm, fetch the prebuilt bundle from GitHub Releases:

curl -fsSL https://raw.githubusercontent.com/nathanaronson/hp-sp-26/main/install.sh | bash

Both install the same self-contained Node bundle (built by CI from cli/). Requires Node ≥ 20. Point at a backend with export DPLOY_API_URL=https://your-backend.

Getting Started (development)

Prerequisites

• uv — Python package manager
• pnpm — Node package manager

Install dependencies

make install

Run the full stack (backend + frontend)

make dev

This starts the FastAPI backend on http://localhost:8000 and the Vite frontend dev server concurrently. Ctrl-C stops both.

Run individually

make backend    # FastAPI only
make frontend   # Vite only

Run the CLI

make cli ARGS="deploy https://github.com/org/repo"

Pass any dploy arguments via ARGS. See dploy --help for the full command reference.

CLI Commands

dploy deploy [path|github-url]   Deploy a local project or GitHub repo
dploy list                       List all deployments
dploy status <id>                Check deployment status
dploy stop <id>                  Tear down a deployment
dploy open <id>                  Open the deployment URL in your browser
dploy login                      Authenticate via GitHub OAuth
dploy logout
dploy whoami                     Print the current user

Other Make Targets

make test       # Run backend tests (pytest)
make lint       # Lint backend (ruff) + frontend (eslint)
make fmt        # Format backend with ruff
make api        # Regenerate typed API client from OpenAPI schema
make build      # Production build of the frontend
make clean      # Remove caches, dist artifacts, and local DB

How It Works

Each deployment follows this pipeline:

1. Provision or acquire a warm sandbox (pool kept ready to avoid cold-start latency)
2. Clone the repo / receive the uploaded tarball
3. Run a deterministic heuristic analysis — common stacks skip the LLM entirely
4. If needed, Agent #1 inspects the codebase and produces an install/build/start plan
5. CLI repos are exposed as a browser terminal via ttyd
6. Web apps go through Agent #2: build, start, discover ports, verify HTTP, report live services
7. Tunnels are opened for detected ports and URLs are stored

Authors

• Nathan Aronson
• Kaitlyn Kwan
• Ryan Tanenholz
• Samuel Lao