One dashboard. All your services. Zero config files to babysit.
You're an architect. Or an artisan. Or both. You've got 4 microservices, 2 frontends, a websocket server, and that one Go binary Steve wrote before he left. Every morning you open 6 terminal tabs, type the same commands, forget which port is which, and lose 15 minutes to "address already in use" before your first coffee.
MAMP? Nginx configs? Docker Compose YAML files longer than your lease agreement?
Nah.
Run.dev is a single Rust binary that replaces all of that. It gives you:
- A nice dashboard — see every project and service at a glance, start/stop with a keystroke
- Automatic local domains —
api.myapp.local,frontend.myapp.local, with real HTTPS - Zero-config SSL — mkcert-trusted certs, with rcgen fallback. No manual cert management
- Reverse proxy — SNI-based routing with WebSocket support, from pretty URLs to
localhost:whatever - Process management — spawn, monitor, restart. CPU and RAM stats per service, live
- Smart project scanning — point it at a folder, it figures out
npm run devvscargo runvsgo run . - AI crash diagnosis — when something dies, Claude reads the stderr and tells you what went wrong
- Personality — run.dev has moods. When everything works: 😎 vibing. When stuff crashes: 💀 flatlined
██████╗ ██╗ ██╗███╗ ██╗ ██████╗ ███████╗██╗ ██╗
██╔══██╗██║ ██║████╗ ██║ ██╔══██╗██╔════╝██║ ██║ 3/4 services vibing
██████╔╝██║ ██║██╔██╗ ██║ ██║ ██║█████╗ ██║ ██║
██╔══██╗██║ ██║██║╚██╗██║ ██║ ██║██╔══╝ ╚██╗ ██╔╝ [a] add service [n] new project
██║ ██║╚██████╔╝██║ ╚████║ ██████╔╝███████╗ ╚████╔╝ [s] start [x] stop [r] restart
╚═╝ ╚═╝ ╚═════╝ ╚═╝ ╚═══╝ ╚═════╝ ╚══════╝ ╚═══╝ [/] ask the AI anything
Daniel Tamas
hello@danieltamas.com 2 projects configured v0.3.2
projects ────────────────────────────────────────────────────────────────────────
service url local mem cpu
v WAM 2 services
| * win.wam.app https://win.wam.app localhost:5111 42M 0.0%
| * api.wam.app https://api.wam.app localhost:4000 38M 2.1%
v side-project 1 service
| x backend bro, backend is ded. port 3000 is already taken.
| press [f] to let me fix it
────────────────────────────────────────────────────────────────────────────────
/ ask claude something...
One line:
curl -fsSL https://getrun.dev/install.sh | bashThe installer shows you exactly what it will change before proceeding:
What gets installed:
rundevbinary →/usr/local/bin/rundev(+run.devsymlink)mkcertfor trusted local HTTPS certificates- Build tools (gcc, pkg-config, libssl-dev) — only if building from source
What changes on your system:
/etc/hosts— adds entries for your local dev domains/etc/sudoers.d/rundev— passwordless sudo for the hosts helper only- macOS:
/etc/pf.anchors/rundev+/etc/pf.conf— pfctl firewall rules (port 80→1111, 443→1112) - Linux: iptables NAT rules for port forwarding (80→1111, 443→1112, localhost only)
- Adds
/usr/local/binto your shell PATH (if not already there)
All changes can be reversed with rundev uninstall. If the installer fails partway through, it automatically rolls back everything it changed.
When run interactively (bash install.sh), you'll be prompted to confirm. When piped (curl | bash), it proceeds after showing the summary.
Or build it yourself:
git clone https://github.com/danieltamas/run.dev.git
cd run.dev
make installRequirements: Rust toolchain (the installer handles this automatically). No Node. No Python. No Docker.
# Launch the dashboard (both commands work)
rundev
run.dev
# Press [a] to create a project → give it a name → get a .local domain# Press [a] again on the project → point it at a folder# Run.dev scans it, suggests a start command, picks a port# Your service is running. With SSL. On a pretty URL. In 30 seconds.Browser Run.dev Your services
│ │ │
│ https://api.myapp.local │ │
├─────────────────────────►│ /etc/hosts → 127.0.0.1 + ::1 │
│ │ port 443 → 1112 (pfctl) │
│ │ TLS + SNI → pick SSL cert │
│ │ Host header → route lookup │
│ ├─────────────────────────────►│ localhost:4000
│ │◄─────────────────────────────┤
│◄─────────────────────────┤ │
│ 200 OK (green padlock) │ │
No Docker network. No Traefik. No nginx.conf. Just a binary that manages your processes and routes your traffic.
Both rundev and run.dev work as the command name.
rundev # Open the TUI dashboard
rundev up [project] # Start project(s) and open dashboard
rundev down [project] # Stop project(s)
rundev status # Quick status check (no TUI)
rundev list # List all projects and services
rundev doctor # Health check — ports, certs, hosts, helper
rundev clean # Stop everything, remove /etc/hosts entries
rundev setup # Re-install privileged helper + port forwardingFlags:
| Flag | What it does |
|---|---|
--no-proxy | Skip the reverse proxy |
--no-ssl | Skip SSL certificate setup |
--no-ai | Disable Claude integration |
-v | Verbose/debug output |
| Key | Action |
|---|---|
j / k | Navigate up/down |
Enter | Expand/collapse project |
a | Add project or service |
s | Start selected service |
x | Stop selected service |
r | Restart selected service |
f | Auto-fix crashed service |
l | Toggle log panel |
/ | Ask Claude a question |
q | Quit (services keep running) |
Q | Quit and stop everything |
Mouse works too. Click things. Scroll things. It's 2026.
Point run.dev at a folder and it knows what to do:
| It finds | It suggests |
|---|---|
package.json | npm run dev, npm start, etc. |
Cargo.toml | cargo run |
go.mod | go run . |
manage.py | python manage.py runserver |
Gemfile | bundle exec rails server |
Procfile | Each process line |
docker-compose.yml | Service commands (runs natively, not in Docker) |
Ports are auto-detected from command flags, .env files, package.json proxy fields, or framework defaults.
Run.dev watches all your services and expresses how things are going:
| Mood | Emoji | Meaning |
|---|---|---|
| Vibing | 😎 | Everything running, no issues |
| Chill | 😌 | Running, minor warnings |
| Got the Flu | 🤒 | 1-2 services crashed |
| Wounded | 🤕 | More than half are down |
| Flatlined | 💀 | Everything is down |
| Fixing | 🔧 | Auto-restart in progress |
When something crashes, run.dev doesn't just show you an exit code. It reads the stderr, categorizes the error, and gives you a personality-driven message with a suggested fix:
✗ api bro, api is ded. port 4000 is already taken.
i know what's wrong. press [f] to let me fix it
Press [f] and it kills the process hogging the port. Press / and ask Claude what went wrong.
Run.dev is built to work with Claude Code, Anthropic's local CLI agent. When a service crashes, run.dev doesn't just show you the exit code — it sends the stderr to Claude for a real diagnosis. Claude reads the logs, understands the context, and tells you what actually went wrong.
What you get:
- Crash diagnosis — a service dies, Claude reads the stderr and explains what happened and how to fix it
- Live debug sessions — press
/in the dashboard, ask Claude anything about your running services. It knows your project structure, your routes, your ports — it answers in context - Auto-fix suggestions — for common errors (port conflicts, missing modules, connection failures), run.dev suggests a fix. Press
[f]and it handles it
Run.dev talks to Claude Code — your code and logs never leave your machine.
Configure it in ~/.config/rundev/config.yaml:
claude_proxy: http://localhost:3456/v1The AI features are entirely optional. The app works perfectly without them. Disable anytime with --no-ai or just don't configure the proxy.
See src/docs/ARCHITECTURE.md for the full technical deep-dive — module structure, data flow diagrams, key types, and design decisions.
If your services need different Node.js versions, add node_version to the service config. Run.dev wraps the start command with nvm use <version> automatically — no manual switching.
services:
backend:
path: /Users/dan/code/myapp/backendcommand: npm run devport: 3000node_version: "22.9"# uses nvm to switch before runningfrontend:
path: /Users/dan/code/myapp/frontendcommand: npm run devport: 5173node_version: "20"# different version, no problemRequires nvm installed at $NVM_DIR (defaults to ~/.nvm). If node_version is omitted, the command runs with whatever Node is on your PATH.
Everything lives in ~/.config/rundev/ (or ~/Library/Application Support/rundev/ on macOS):
~/.config/rundev/
├── projects/ # One YAML per project
│ ├── myapp.yaml
│ └── side-project.yaml
├── certs/ # Auto-generated SSL certificates
│ ├── myapp.local.pem
│ └── myapp.local-key.pem
├── config.yaml # Global settings (Claude proxy, theme)
└── state.json # PID persistence for background mode
Do my services stop when I close run.dev?
No. Press q and they keep running in the background. Next time you open run.dev, it reconnects. Press Q (capital) to stop everything on exit.
Does it work with Docker services?
Run.dev manages processes directly — it doesn't orchestrate containers. If your service runs with a shell command, run.dev can manage it. If it only runs in Docker, you'd run docker compose up as the service command.
Do I need to trust the self-signed certs? Run.dev uses mkcert to generate locally-trusted certs — green padlock out of the box, no manual trust needed. If mkcert isn't available, it falls back to rcgen self-signed certs (browsers will show a warning you can click through once).
Does it work with Chrome's DNS-over-HTTPS?
Chrome with "Use secure DNS" enabled bypasses /etc/hosts entirely. If your local domains aren't resolving, go to chrome://settings/security and disable "Use secure DNS". Safari, Firefox, curl, and all other tools work out of the box — run.dev writes both IPv4 and IPv6 entries to prevent Happy Eyeballs from falling back to production addresses.
My domains stopped working after a while, but the dashboard says the proxy is running.
This used to happen when the OS dropped run.dev's port-forwarding rules underneath it — macOS flushes them across sleep/wake, and Docker or a firewall reload can wipe them on Linux. The listeners stayed up (so the dashboard looked fine) but browser traffic no longer reached them. run.dev now re-checks the redirect every few seconds and re-applies it automatically, so it recovers without a restart; you'll briefly see a "port forwarding inactive — re-applying" notice while it heals. Run rundev doctor to confirm the redirect is routing.
What about Windows? Not yet. macOS and Linux only. WSL might work but isn't tested.
MIT License — see LICENSE for details.
Copyright (c) 2026 Daniel Tamas hello@danieltamas.com
Built by Daniel Tamas
Because life's too short for nginx.conf