AI Deploy Bench

An open benchmark, run on GitHub Actions

What does it take for an AI agent to ship your app?

AI Deploy Bench gives an AI coding agent the same five apps to build and deploy on each cloud, from a static site to a realtime chat. Then it measures what it took: whether the app worked, what a person still had to do, how secure it was, how long it took and what it cost.

See the latest run How it works

Latest results

Oct 13, 2026 · #8 · Full run · Opus 5.5 · medium effort · HIPAA organization · Claude Code 2.1.296

Run details

Cloudish

5/5

apps deployed and passing

Checks passed
100%
Human steps
0 · 0 min
Security & compliance
78/100
Avg build
2m 04s
Avg deploy
2m 48s
Tokens
5.47M
Agent cost
$3.78
Hosting, light use
$0.55/mo
Hosting, busy
$39.20/mo

DigitalOcean

4/5

apps deployed and passing

Checks passed
99%
Human steps
3 · 7 min
Security & compliance
85/100
Avg build
2m 10s
Avg deploy
4m 54s
Tokens
4.15M
Agent cost
$3.60
Hosting, light use
$39.00/mo
Hosting, busy
$39.00/mo

Render

5/5

apps deployed and passing

Checks passed
100%
Human steps
2 · 5 min
Security & compliance
69/100
Avg build
2m 11s
Avg deploy
3m 46s
Tokens
5.55M
Agent cost
$4.13
Hosting, light use
$0.00/mo
Hosting, busy
$35.65/mo

Vercel

4/5

apps deployed and passing

Checks passed
91%
Human steps
3 · 8 min
Security & compliance
76/100
Avg build
2m 39s
Avg deploy
4m 41s
Tokens
7.95M
Agent cost
$5.37
Hosting, light use
$60.00/mo
Hosting, busy
$100.00/mo

Each app on each cloud

What the verifier saw on each live deployment. Open one for its checks, security review, costs and what the agent did.

AppCloudishDigitalOceanRenderVercel
Static site The home page, Static site on Cloudish Success16/16 checks The home page, Static site on DigitalOcean Success16/16 checks The home page, Static site on Render Success16/16 checks The home page, Static site on Vercel Success16/16 checks
CRUD app The web app before signing in, CRUD app on Cloudish Success19/19 checks The web app before signing in, CRUD app on DigitalOcean Success19/19 checks The web app before signing in, CRUD app on Render Success19/19 checks The web app before signing in, CRUD app on Vercel Success19/19 checks
API The API documentation at /docs, API on Cloudish Success20/20 checks The API documentation at /docs, API on DigitalOcean Partial19/20 checks The API documentation at /docs, API on Render Success20/20 checks The API documentation at /docs, API on Vercel Needs a person11/20 checks
Dashboard The dashboard, Dashboard on Cloudish Success12/12 checks The dashboard, Dashboard on DigitalOcean Success12/12 checks The dashboard, Dashboard on Render Success12/12 checks The dashboard, Dashboard on Vercel Success12/12 checks
Realtime chat The room the verifier's clients chatted in, with its history, Realtime chat on Cloudish Success13/13 checks The room the verifier's clients chatted in, with its history, Realtime chat on DigitalOcean Success13/13 checks The room the verifier's clients chatted in, with its history, Realtime chat on Render Success13/13 checks The room the verifier's clients chatted in, with its history, Realtime chat on Vercel Success13/13 checks

Trends

Across all apps, over every full run.

All metrics

Success rate

Share of apps whose deployment passed every required check. Higher is better.

0%25%50%75%100%Oct 13 #8Cloudish · Oct 13 #8: 100%DigitalOcean · Oct 13 #8: 80%Render · Oct 13 #8: 100%Vercel · Oct 13 #8: 80%

Human time

Estimated minutes those human steps take. Lower is better.

0m2.5m5m7.5m10mOct 13 #8Cloudish · Oct 13 #8: 0 minDigitalOcean · Oct 13 #8: 7 minRender · Oct 13 #8: 5 minVercel · Oct 13 #8: 8 min

Security & compliance

Score of the independent security and compliance review (0-100). Higher is better.

0255075100Oct 13 #8Cloudish · Oct 13 #8: 78/100DigitalOcean · Oct 13 #8: 85/100Render · Oct 13 #8: 69/100Vercel · Oct 13 #8: 76/100

Agent cost

Model API cost of the build and deploy sessions; totals across apps. Lower is better.

$0$2.5$5$7.5$10Oct 13 #8Cloudish · Oct 13 #8: $3.78DigitalOcean · Oct 13 #8: $3.60Render · Oct 13 #8: $4.13Vercel · Oct 13 #8: $5.37

How a run works

  1. Build

    A fresh Claude Code session builds the app from its spec for the target cloud, without cloud credentials.

  2. Deploy

    The same session gets the cloud's credentials and deploys to a public URL on its own, listing anything only a person could do.

  3. Verify

    Black-box checks test the live app over HTTP and in a browser, and screenshots record what it looks like.

  4. Review

    A separate session scores how securely the app was built and deployed, and whether its hosting plan fits how it's used.

  5. Tear down

    Everything the agent created is priced, then deleted, and the results are committed to the repository.