Buying snapshot
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Decision point | What the Berth project states |
|---|---|
License | Apache License 2.0 |
Where code runs | Your Mac, or Linux servers running Berth's single Rust agent binary (systemd service). AWS Lambda and Cloudflare Workers are Phase 4 targets, marked Not Started. |
Setup | App Quick Start: cargo tauri dev, with Rust, Node.js and protoc as prerequisites. Linux agent: one-line install script (curl ... | sudo bash). |
Agent connection | Synadia Cloud (recommended): zero inbound ports, works behind NAT, needs a free account and a .creds file. Direct Connection: mTLS over gRPC, needs network reachability. |
Supported code | Python, Node, Go, Rust, Shell and Docker Compose; parses requirements.txt, package.json, go.mod, Cargo.toml and docker-compose.yml |
Interfaces | Desktop GUI, a CLI with full command parity, and an MCP server over JSON-RPC 2.0 stdio. The README counts 19 MCP tools; the technical documentation lists 26. |
Our take
Berth is worth a trial if you are a developer or small team whose AI-written scripts, bots and services should run on machines you already own, with Claude Code driving deployment over MCP. Start with jobs you can afford to lose: much of the roadmap, from cloud targets to team features, is still ahead.
Evaluate Berth
Comparisons
Where it stands out
Remote servers without open ports. With NATS (Synadia Cloud), neither the desktop nor the agent exposes ports, and it works behind NAT, firewalls and across networks. gRPC over HTTP/2 on port 50051 is the fallback. Servers in a home lab or behind a strict firewall stay reachable.
Jobs keep running without the app. Scheduling takes @every 5m, @hourly or expressions like 30 9 * * *. The agent-side scheduler checks every 30s and runs jobs even when the app is offline.
Long-running services. Service Mode keeps API servers, bots and workers up, with auto-restart on crash and exponential backoff. Agent self-upgrade swaps the binary atomically, then restarts under a 30s probation with automatic rollback if the new agent fails it.
Public URLs from your own machine. Publish gives a running service a public URL through Cloudflare Quick Tunnel (free), ngrok or a custom command. Tunnels run on your machine, with zero Berth-side infrastructure.
Less setup per project. Deploys to a remote agent auto-run pip install, npm install or go mod download. Each project has its own environment variables, with .env import and secret masking in logs.
Live logs. stdout and stderr stream into an xterm.js terminal with ANSI colors and 10k lines of scrollback.
Product screenshot
Trade-offs to weigh
Mac-only control plane. Windows and Linux builds sit in the roadmap's Phase 5, next to team features and a web dashboard. A team with mixed desktops cannot give everyone the app today.
Homebrew cask is on the roadmap. The roadmap lists a Homebrew cask in Phase 4, which is Not Started. Setup follows the Quick Start prerequisites in the snapshot, so plan for someone comfortable with Rust tooling.
MCP access is deployment access. MCP tools include berth_deploy, berth_publish, berth_env_set and berth_add_target. Any MCP client you connect can ship code, open a public URL or change environment variables, so review that access like any deploy key.
berth_logs is only partly done. The technical documentation marks the berth_logs MCP tool (fetch logs, pending log store) as Partial. Test it before your workflow depends on it.
Pricing and plan conditions
Berth is open source under Apache 2.0. You pay for the Mac and the Linux servers it runs on. Public URLs use cloudflared by default, and Cloudflare Quick Tunnel is labelled free. The recommended Synadia Cloud connection mode needs a free Synadia account. The roadmap lists a Pro tier under Phase 5: Growth.
Public signals
About these signals
Verified factual signals from public sources. They indicate observable activity or interest, not total adoption, product quality, or cost.
0 GitHub commits 90d4 GitHub stars
See all signals from 3 sources
Alternatives to Berth
The reviewed substitutes for Berth among the container platforms, and what would make each one the better answer.
Other approaches
A different approach to the same problem. Each substitutes only for the workload named beside it.
- Kubernetes
- Kubernetes schedules and scales containers across a cluster; Berth keeps services and scheduled jobs running on individual Mac or Linux servers with no cluster and no YAML. A team with a few small services can use Berth instead; Kubernetes is the choice once those services must scale across many machines.Applies to: Keeping a few small services running on servers you own.
- Docker
- Pick Docker when an app must move between machines or be handed to another team.Applies to: Running a small app or scheduled job on a Mac or Linux server you own.
- Nomad
- Nomad and Berth both run services, binaries and scheduled jobs on your own servers through a single agent binary. Nomad schedules them across a cluster from job files; Berth deploys code from a Mac app, a CLI or an MCP server.Applies to: Running services and scheduled jobs on your own servers without Kubernetes.
Related technologies
Normally used together rather than chosen between, so these are not alternatives.
- Cursor
- Cursor writes code and Berth runs it: Berth's documentation names code written in Cursor, Claude Code or Codex as what it deploys to your own Mac and Linux servers. Neither replaces the other.Applies to: Writing an app with AI in Cursor, then running it on your own servers with Berth.