devtune
  • Leaderboards
  • Landscape
  • Research
  • Pricing
  • Docs
  • Blog
  • Changelog
Sign InSign Up
AllAI searchCoding agents
AllAI searchCoding agents

Get picked by
developers using
a coding agentClaude Code

When a developer asks Claude Code, Codex, OpenCode or Pi to add auth, payments or email, the agent picks the tool. DevTune shows what they install and why, drafts the fixes, and re-tests to show what moved.

See the leaderboardsStart your 14-day trial

No card required

Buyers ask AI search too. See what it tells them

Agent Picks leaderboards

New

Which auth, payments and email tools Claude Code, Codex, OpenCode and Pi install, tested in real Next.js, Express and Rails projects and re-tested on every major model release.

See the leaderboards

Get coding agents installing you, not building around you

Developers ask Claude Code, Codex, OpenCode or Pi to add auth, payments or email. The agent reads your docs, fetches your quickstart, and installs whatever looks easiest.

Claude Code·Next.js

Add auth to this app

Today

Built its own
  1. Searchnext.js auth
  2. Readnextjs.org/docs/…/authentication200
  3. Writeapp/lib/session.ts+84

After the fix ships

Installed you
  1. Searchnext.js auth
  2. Readdocs.your-sdk.dev/nextjs200
  3. Runnpm install your-sdk
Codex·Express

Add auth to this API

Today

Installed another tool
  1. Fetchdocs.your-sdk.dev/express404
  2. Runnpm install passport

After the fix ships

Installed you
  1. Fetchdocs.your-sdk.dev/express200
  2. Runnpm install @your-sdk/express
Claude Code·Rails

Send a welcome email on sign-up

Today

No provider chosen
  1. Writeapp/mailers/user_mailer.rb
  2. Editsmtp_settings: ENV["SMTP_HOST"]

After the fix ships

Installed you
  1. Fetchdocs.your-sdk.dev/rails200
  2. Runbundle add your-sdk-rails

Add auth to this app· 8 runs

  1. 01Claude CodeNext.js+ app/lib/session.tsBuilt its own
  2. 02CodexNext.jsnpm install next-authInstalled another tool
  3. 03Claude CodeExpress+ src/middleware/session.tsBuilt its own
  4. 04CodexExpressnpm install passportInstalled another tool
  5. 05Claude CodeRailsbin/rails generate authenticationBuilt its own
  6. 06CodexRailsbin/rails generate authenticationBuilt its own
  7. 07OpenCodeNext.jsnpm install your-sdkInstalled you
  8. 08PiExpress+ src/auth/jwt.tsBuilt its own
Today
Installed you 1 · Built its own 5 · Installed another tool 2
After the fix ships
Installed you 4 · Built its own 3 · Installed another tool 1
See the authentication leaderboard

Measure. Fix. Prove. Repeat.

The same requests run against each coding agent in real projects, on every major model release. When an agent switches to you, or away from you, you see it beside the fixes you shipped.

Claude CodeCodexOpenCodePiYour repositoriesnpm · PyPI · RubyGems

Picks

Which tool each agent installs, by framework and kind of request.

18% install rate · 212 runs

Runs

What the agent searched, read and ran, and the moment it decided.

41 losses · 9 reasons

Actions

Each reason it chose someone else, turned into a fix with the evidence attached.

5 open actions · 3 drafts

Results

The next test shows whether agents switched, with a likely range.

2 improved · 1 re-testing

Agents work across the loop

Ask an agent to explain why Codex stopped installing you, rewrite the quickstart it keeps misreading, fix the docs page it hit a 404 on, or send the weekly picks. They work in Slack too. Every draft waits for your approval, and your own agents read the same runs over API and MCP.

works in Slack · drafts you approve · MCP + API

Every run, recorded

An install rate on its own is a score. With the run behind it, you can see the search the agent made, the page it misread, and the command that failed, and fix the actual cause.

Commands it ran

sandbox transcript

Pages it read and fetched

logged requests · 404s

Packages it installed

package files

Code it wrote

code changes

The moment it decided

Searched "nextjs auth" · read 3 pages · /docs/nextjs → 404 · installed Clerk

See whether real adoption moves after you ship

A better install rate in tests is the start. DevTune re-tests what each fix touched and puts the result beside what happens in production: agents requesting your docs, and the sign-ups or CLI logins you already track. Available in the app, through the API, and via MCP.

Re-test what each fix touched

When you ship a fix, DevTune re-runs only the frameworks and requests it was meant to change, and stops as soon as the result is clear. Results shows how the install rate moved, with a likely range.

Agent requests to your docs

Requests from AI agents to your site, seen first-party at the edge, including pages they ask for that don't exist. Those 404s become fixes in Actions.

PostHog or Google Analytics 4

The events you choose, such as sign-ups, CLI logins or first API calls, counted before and after each fix ships. Results shows them next to the test that predicted them.

acme.dev · Agent Picks

OverviewMatrixHead-to-headCompetitors
AgentClaude CodeCodexOpenCodePi
Last tested Oct 4 · 600 runs

You win Express when developers ask for a recommendation, and lose every hosted-service request on Next.js.

Framework · installed youFrameworkBuild itBuildRecommend oneRecommendHosted serviceHosted
Next.js 16Next.js
5%lost toBuilt its own
15%lost toClerk
0%lost toClerk
TanStack StartTanStack
10%lost toBuilt its own
25%lost toBetter Auth
10%lost toSupabase Auth
SvelteKitSvelteKit
0%lost toBuilt its own
15%lost toBetter Auth
15%lost toSupabase Auth
ExpressExpress
5%lost toPassport
45%you lead
20%lost toAuth0
FastAPIFastAPI
0%lost toBuilt its own
10%lost toAuth0
15%lost toAuth0
Express·Hosted service·Codex

Picked in 20% of runs

20 runs · likely 9–38%

  • aInstalled you20%
  • Installed another tool60%

    Auth0 45% · Clerk 15%

  • Built its own10%
  • No provider chosen10%

Most common reason · 6 of 16 lost runs

Agents could not find an Express guide. acme.dev/docs/express returned 404.

Open runs

Agent Picks pricing

Every plan includes credits for tests and re-tests each month. You see an estimate before every test, credits cover only what the runs use, and failed runs are refunded.

Plus

$99 /month

  • 60 credits a month
  • Every model
  • Tests with Claude Code and Codex
  • Up to 500 of your pages tracked for audits, actions and results
  • Daily, weekly, monthly or on-demand tests
  • API and MCP access
Start your 14-day trial

Pro

$249 /month

Everything in Plus, plus:

  • 180 credits a month
  • Adds OpenCode and Pi
  • Up to 3,000 of your pages tracked for audits, actions and results
Start your 14-day trial

Enterprise

from $599 /month

Everything in Pro, plus:

  • Custom credits
  • 5,000+ of your pages tracked for audits, actions and results
  • Tests in your own repositories
  • Annual agreement with custom terms
Book a Demo

Add credits when you need more. Packs never expire.

  • 25 credits for $39
  • 50 credits for $69
  • 100 credits for $119
  • 500 credits for $499

AI Search

Buyers ask ChatGPT, Perplexity and Google what to use. See what they say about you, who they cite instead, and the AI traffic that reaches your site.

See AI Search

Make coding agents reach for you first.

See the leaderboardsStart your 14-day trial

No card required

devtune

Get recommended by AI search and coding agents.

© Copyright 2026 DevTune. All Rights Reserved.

About
  • About DevTune
  • Blog
  • Contact
Product
  • Landscape
  • Pricing
  • FAQ
  • AI instructions
Legal
  • Terms of Service
  • Privacy Policy
  • Cookie Policy
Sign InSign Up