Blacksmith vs Depot
Blacksmith and Depot are both drop-in GitHub Actions runners: you change one label and each job gets a fresh machine that is thrown away when it ends. Blacksmith runs each job in a Firecracker microVM on its own hardware at a lower per-minute rate, while Depot runs each job on a new AWS EC2 instance and adds fast Docker builds, egress rules and Claude Code sandboxes.
The companies and products
Blacksmith
Blacksmith is a CI service: it runs your GitHub Actions jobs on its own faster, cheaper runners instead of GitHub's.
You change the runner label in your workflow file, and each job runs in a fresh Firecracker microVM that boots in under 3 seconds and is destroyed when the job ends. It adds Docker layer caching and sticky disks that carry chosen folders from one run to the next, plus a Testbox CLI in early beta that lets a local coding agent run commands inside a CI job. It is built for teams that want quicker, cheaper GitHub Actions, not as a general sandbox you drive by API; a separate Sandboxes product is marked coming soon.
Depot
Depot is a build and CI platform: faster Docker builds, drop-in GitHub Actions runners, its own Depot CI, and newer sandboxes for coding agents.
Depot's main product is GitHub Actions runners: you change one runs-on label and each job runs on a fresh AWS EC2 instance that is thrown away when the job ends, with Linux, Arm, Windows and macOS sizes and a shared fast cache. It also sells remote agent sandboxes that run Claude Code in a container, and a Sandbox SDK in public beta whose sandboxes live at most 24 hours. It is built for engineering teams that want CI and builds to finish sooner, so it is a different kind of product than an agent sandbox, and this page compares where the two overlap.
Which use cases each one fits
For each use case, Battleships prices every sandbox provider on the same workload and sorts them from the cheapest monthly bill up. “78th cheapest of 377” means 77 of the other 376 providers run that workload, with every feature it needs, for less. “With limits” means the provider runs it but misses something, so it is listed after every full match. Open a use case for its whole ranking and what each provider misses.
| Use case and its workload | Blacksmith: price rank | Depot: price rank | boat: price rank |
|---|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Agents that build and run whole apps need hosting, a kept disk and forks, which boat has; Blacksmith and Depot cover only the CI part of that loop. | 248th of 367, with limits $3,607 a month | 331st of 367, with limits $8,688 a month | 3rd cheapest of 367 $634 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that lives on one machine all month needs a disk that survives and a way to reach it; Blacksmith and Depot runners end with the job, and Depot's Sandbox SDK caps a sandbox at 24 hours. | 78th of 366, with limits $5,463 a month | 86th of 366, with limits $13,148 a month | 5th cheapest of 366 $657 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month | 281st of 345, with limits $488 a month | 311th of 345, with limits $1,188 a month | 186th of 345, with limits $2,000 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Giving each of your users a sandbox means starting machines from your own backend and reaching them by URL, which boat does by API and neither CI runner does. | 268th of 361, with limits $1,563 a month | 318th of 361, with limits $3,713 a month | Cheapest of 361 $209 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 276th of 360, with limits $7,488 a month | 311th of 360, with limits $17,908 a month | 34th of 360, with limits $2,000 a month |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month Agent evals need many clean machines you start from code; boat starts and forks them by API, while both CI runners start machines only from workflow jobs. | 354th of 364, with limits $1,188 a month | 319th of 364, with limits $2,888 a month | 10th cheapest of 364 $200 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month Both are built for this and both fit; Blacksmith charges less per runner minute, Depot adds fast Docker builds and egress rules, and boat fits CI you drive by API on a VM that stays up. | 78th cheapest of 377 $708 a month | 125th cheapest of 377 $1,736 a month | 4th cheapest of 377 $86 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Both throw away the machine when a job ends, so a dev box that keeps its files and stays reachable all month fits boat better. | 180th of 368, with limits $207 a month | 199th of 368, with limits $534 a month | 18th cheapest of 368 $26 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Computer-use agents need a screen; boat streams a 60fps Linux desktop, while Blacksmith and Depot runners are headless. | 243rd of 369, with limits $1,088 a month | 321st of 369, with limits $2,600 a month | 6th cheapest of 369 $200 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month | 131st of 367, with limits $588 a month | 139th of 367, with limits $1,304 a month | 7th cheapest of 367 $200 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month | 130th of 346, with limits $719 a month | 139th of 346, with limits $1,763 a month | 6th cheapest of 346 $90 a month |
| Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month | 262nd of 338, with limits $14,613 a month | 312th of 338, with limits $35,108 a month | 71st of 338, with limits $1,800 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month | 64th of 360, with limits $350 a month | 81st of 360, with limits $863 a month | 2nd cheapest of 360 $200 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month | Can't run it | Can't run it | Can't run it |
| macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month Both rent macOS runners by the minute and boat runs Linux only; Depot ranks 10th here and Blacksmith 28th, because Depot's macOS minutes cost less in this scenario. | 28th of 36, with limits $1,508 a month | 10th of 36, with limits $248 a month | Can't run it |
Key features
| Feature | Blacksmith | Depot | boat |
|---|---|---|---|
| What it is | Drop-in GitHub Actions runners; each machine lives for one workflow job | Drop-in GitHub Actions runners, Depot CI and Docker builds, plus Claude Code sandboxes and a beta Sandbox SDK | A full VM per agent that you create, stop, start and fork by API |
| Isolation | Firecracker microVM per job on Blacksmith's own hardware, destroyed after the job | A new AWS EC2 instance per job, destroyed after the job; agent sessions in an isolated container | Full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts |
| Linux runner sizes | Ubuntu on x64 and ARM, up to 32 vCPU / 128 GB on x64 | Linux x64 and Graviton4 Arm, from 2 to 64 CPUs | x64 Linux, up to 16 vCPU / 32 GB on the $100 plan and above |
| macOS and Windows | macOS on Apple Silicon; Windows Server 2025 in public beta | macOS 26, 15 and 14 at 8 CPU / 24 GB; Windows Server 2025 and 2022 | Linux only; Wine for Windows apps |
| Nested virtualization | Yes on x64 Linux runners, not on ARM | Only on Depot CI sandboxes, not on its GitHub Actions runners; no Hyper-V on Windows | /dev/kvm on every machine, with Docker, root and systemd |
| What carries over between runs | Sticky disks: up to 5 folders per job, evicted after 7 days unused | Runners are never reused; a shared cache; agent sandboxes keep files and the Claude Code conversation | The whole disk, saved every minute and on stop; a stopped sandbox costs nothing |
| How long a machine lives | Until the job ends; Testbox stops after 30 idle minutes by default | Until the job ends; Sandbox SDK default 120 minutes, at most 24 hours | Until you stop it; no session cap on paid plans |
| Fork | No; every machine starts fresh from a job | No fork or clone API | Fork a running sandbox with its disk |
| Build and cache extras | Managed Docker layer cache and sticky disks | Shared cache up to 1000 MiB/s and a separate fast Docker build service | No managed CI cache; caches live on the VM's own kept disk |
| Reaching the machine | Opt-in debug SSH for the user who triggered the job; no ports or URLs | SSH into a running Depot CI job for debugging; no documented inbound ports or URLs | HTTPS URLs for up to 50 ports, token-protected; raw TCP; dedicated IP |
| Coding agents | Its own [code]smith agent from Slack, GitHub or a web UI; Testbox CLI in early beta; Sandboxes coming soon | Claude Code only, started and resumed with the depot claude command | Seven agents preinstalled, including Claude Code and Codex, with a prompt, steer and events API |
| Keeping API keys out of the machine | GitHub Actions secrets and OIDC tokens inside the job | Secrets passed into the job or sandbox as environment variables | Credentials proxy for model providers and any HTTPS API; real keys never enter the sandbox |
| Desktop and computer use | Headless runners | No desktop or computer-use product | 60fps Linux desktop with Chromium and computer use |
| Regions and outbound rules | US east, US west, Amsterdam and Frankfurt; no egress allowlist | US, or the EU if a project chooses it; egress filtering by IP, CIDR or hostname on Linux | EU only (Germany, Finland, France); open internet, no provider-side allowlist |
Every feature Battleships tracks (43)
| Feature | Blacksmith | Depot | boat |
|---|---|---|---|
| Full VM (own kernel) | Yes | Yes | Yes |
| Docker inside | Yes | Yes | Yes |
| Nested virtualization | Yes | Yes | Yes |
| Root | Yes | Yes | Yes |
| Systemd | No | No | Yes |
| Desktop GUI | No | No | Yes |
| Browser | Yes | Yes | Yes |
| Computer-use API | No | No | Yes |
| SSH | Yes | Yes | Yes |
| Persistent disk | Yes | Yes | Yes |
| Pause/resume | No | No | Yes |
| Automatic snapshots | No | No | Yes |
| Fork/clone | No | No | Yes |
| Live fork (no pause) | No | No | Yes |
| Memory fork | No | No | No |
| Custom image (Docker or snapshot) | No | Yes | Yes |
| ≥24 h sessions | Yes | Yes | Yes |
| Idle auto-stop | Yes | Yes | No |
| Wake on request | No | No | No |
| HTTPS preview URLs | No | No | Yes |
| Raw TCP inbound | No | No | Yes |
| Public IPv4 | No | No | No |
| Egress allowlist | No | Yes | No |
| Secret proxy | No | No | Yes |
| Secret proxy for any API | No | No | Yes |
| Preinstalled agents | No | Yes | Yes |
| Agent harness API | No | Yes | Yes |
| MCP server | No | No | No |
| GPU | No | Yes | No |
| Windows | Yes | Yes | No |
| Arm64 | Yes | Yes | No |
| Self-hosting | No | No | No |
| BYOC | No | Yes | No |
| Open source | No | No | No |
| Scoped API keys | Not published | Not published | Yes |
| Spend limits | Not published | Yes | Yes |
| Usage per sandbox (API) | Yes | No | Yes |
| Audit logs | Not published | Not published | No |
| SSO | No | Yes | No |
| SOC 2 | Yes | Yes | No |
| HIPAA | No | Not published | No |
| EU data residency | Yes | Yes | Yes |
| Zero data retention | Not published | Not published | Yes |
Performance
There is no comparable measurement for Blacksmith or Depot: neither was part of the same CPU benchmark run, so we do not compare their speed. boat did 26.69 runs per second on the StarSling hpc-sandbox-benchmarks Node.js loop at 4 vCPU / 8 GB, on dedicated AMD Ryzen 9 9950X hosts. Blacksmith says its runners use fast gaming CPUs, and Depot runs on AWS EC2, but neither has a number measured the same way.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Blacksmith and Depot have no published run in this suite yet, so there is no like-for-like number to compare.
Pricing in five scenarios
Each bill is priced from the providers' public price lists by the open Battleships engine, with the cheapest plan that runs the scenario.
| Scenario | Blacksmith, a month | Depot, a month | boat, a month |
|---|---|---|---|
| One coding agent, workdays 4 vCPU / 8 GB, 50 GB disk, 8 hours a day, 22 days a month, files kept between days | $36 · Pay-as-you-go · with limits | $122 · Developer · with limits | $20 · $20/mo |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $207 · Pay-as-you-go · with limits | $534 · Developer · with limits | $26 · $20/mo |
| 50 agents in parallel 50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside | $2,364 · Pay-as-you-go · with limits | $5,710 · Developer · with limits | $317 · $20/mo × 3 seats |
| Sandboxes for your users 2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once | $663 · Pay-as-you-go · with limits | $1,628 · Developer · with limits | $200 · $100/mo × 2 seats |
| 1,000 short jobs at once 2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month | $1,488 · Pay-as-you-go · with limits | $3,588 · Developer · with limits | $2,000 · $2000/mo · with limits |
Blacksmith: Blacksmith bills per minute of job time, scaled by vCPU and OS: Ubuntu x64 is $0.004 a minute for 2 vCPU, ARM $0.0025, Windows $0.008 and macOS $0.08 for 6 vCPU, with 3,000 free minutes a month. Sticky disks cost $0.50 per GB a month and a static IP $100 a month.
Depot: Depot plans cost $20 or $200 a month and include GitHub Actions, Depot CI and Docker build minutes; after that a 2-CPU Linux runner costs $0.006 a minute and bigger sizes multiply it. Depot CI bills per second per vCPU, and agent sandboxes cost $0.01 a minute.
Depot: Plans start at $20 a month, and the whole fee is usage credit spent per second while a sandbox runs. A stopped sandbox costs nothing and keeps its disk.
Limits in those scenarios
None of the agent scenarios fit Blacksmith or Depot cleanly, because both are CI runners: a machine exists only while a GitHub Actions job runs. On Blacksmith only sticky disk folders survive, and Depot's Sandbox SDK ends a sandbox after 24 hours, so an agent that never stops needs restarts. Both say they have no concurrency limits, though large fleets still have to start as workflow jobs. Blacksmith's prices here use its ARM runners and Depot's its 4-CPU / 16 GB Linux runner, which is part of why Blacksmith comes out lower, for example $207 against $534 a month for one always-on machine.
| Scenario | Blacksmith | Depot | boat |
|---|---|---|---|
| One coding agent, workdays |
|
| None that bite |
| One agent that never stops |
|
| None that bite |
| 50 agents in parallel |
|
| None that bite |
| Sandboxes for your users |
|
| None that bite |
| 1,000 short jobs at once |
|
|
|
Which one should you use?
Use Blacksmith when
you want cheaper, faster GitHub Actions minutes, sticky disks for big folders, or KVM on x64 runners.
- Lower per-minute rates. Teams with a lot of Linux CI minutes pay less: a 2 vCPU Ubuntu x64 runner is $0.004 a minute, against Depot's $0.006.
- Sticky disks and nested virtualization. Useful if jobs reuse big folders or run emulators: sticky disks carry chosen folders forward, and x64 runners expose KVM.
- Two US regions and free minutes. It runs in US east and US west as well as Amsterdam and Frankfurt, with 3,000 free minutes a month.
Use Depot when
you also want fast Docker builds, runners up to 64 CPUs, egress rules, or several macOS and Windows versions.
- Fast Docker builds alongside CI. Useful if image builds are your slow step: Depot speeds up Docker builds as well as GitHub Actions jobs.
- Wider runner range. It offers Linux runners up to 64 CPUs, several macOS versions and Windows Server 2025 and 2022 out of beta.
- Egress rules and Claude Code sandboxes. Useful if jobs must only reach chosen hosts, or you want hosted Claude Code sessions next to your CI.
Where boat fits
Blacksmith and Depot are CI services, and for GitHub Actions jobs one of them is the right pick. If what you are really running is agents, a machine that ends with each job works against you. boat gives each agent a full VM you start by API, keep as long as you like, fork, and reach from outside.
A machine that stays
On Blacksmith and Depot the machine is destroyed when the job ends, so an agent reinstalls its tools every run. Blacksmith's workaround is sticky disk folders at $0.50 per GB a month; Depot's Sandbox SDK still ends after 24 hours.
Apps and screens you can reach
Neither CI runner exposes ports or URLs, only SSH for debugging, so you cannot open what an agent built or give a computer-use agent a screen. boat serves HTTPS on 50 ports, raw TCP and a 60fps desktop.
Agents driven from outside, keys kept out
Blacksmith machines start from GitHub workflows and Depot only runs Claude Code from its CLI, with secrets as environment variables any code can read. boat lets you drive agents inside from outside by API, with real keys held in a credentials proxy.
Use boat when what you are running is agents, not CI jobs: a full VM you start by API that keeps its disk, forks, hosts what it builds and runs coding agents you drive from outside. Compare it one to one: boat vs Blacksmith and boat vs Depot.
Questions people ask
Is Blacksmith cheaper than Depot?
Per minute of Linux CI, yes: a 2 vCPU Ubuntu x64 runner is $0.004 a minute on Blacksmith and a 2-CPU Linux runner $0.006 on Depot. In our scenarios Blacksmith also comes out lower, for example $2,364 against $5,710 a month for 50 machines, partly because it is priced on ARM runners. For macOS minutes, Depot came out cheaper.
Which is faster, Blacksmith or Depot?
We have no comparable CPU measurement for either, so we do not rank them on speed. Blacksmith runs on its own hardware with gaming CPUs, and Depot runs each job on AWS EC2 with a shared cache up to 1000 MiB/s.
Can Blacksmith and Depot run Docker?
Yes, both run Docker on their Linux runners. Blacksmith adds a managed Docker layer cache, and Depot has a separate fast Docker build service. Nested virtualization differs: Blacksmith exposes it on x64 Linux runners, while Depot offers /dev/kvm only on Depot CI sandboxes, not on its GitHub Actions runners.
Can I run an AI agent on Blacksmith or Depot?
Partly. Depot sells remote sandboxes that run Claude Code and a Sandbox SDK in public beta whose sandboxes live at most 24 hours. Blacksmith has a Testbox CLI in early beta and its own [code]smith agent, with Sandboxes marked coming soon.
Do both have macOS and Windows runners?
Yes. Blacksmith runs macOS on Apple Silicon and Windows Server 2025 in public beta; Depot runs macOS 26, 15 and 14 and Windows Server 2025 and 2022. boat runs Linux only.
Is boat SOC 2 compliant?
SOC 2 is in progress. We're open about where we stand: live status at status.ascii.dev and our data retention policy at docs.boat.dev/data-retention.
Try boat
Install the CLI, then run boat new.
7-day free trial