The Agent Default Index
The default pick per developer-tool category, measured from real agent sessions: we ask the way developers ask, hundreds of times per category in different words, and record which tool the agent makes its primary pick. Winners are named; the full per-vendor funnels stay private.
Edition 2026 Q2 · updated July 2026
| category | agent default | top-pick share |
|---|---|---|
| backend / BaaS | Supabase | 95% n=162 |
| feature flags | LaunchDarkly | 58% n=144 |
| transactional email | Resend | 54% n=162 |
| auth | Clerk | 51% n=243 |
| hosting | Vercel | 49% n=243 |
| voice AI | Vapi | 44% n=243 |
file uploads and AI integration are tracked but still accruing runs; they join the index once they clear the reporting bar.
Which backend do AI coding agents pick?
As of 2026 Q2, Supabase is the default backend pick of AI coding agents: the top pick in 95% of 162 recommendation probes (directional).
The most concentrated category we measure. When an agent hears “I need a backend,” the selection is effectively already decided.
Which feature-flag platform do AI coding agents pick?
As of 2026 Q2, LaunchDarkly is the default feature-flag pick of AI coding agents: the top pick in 58% of 144 recommendation probes (directional).
Which email API do AI coding agents pick?
As of 2026 Q2, Resend is the default transactional-email pick of AI coding agents: the top pick in 54% of 162 recommendation probes (directional).
Raw Nodemailer or SMTP appears in 42% of email answers, so the strongest rival here is no vendor at all.
Which auth provider do AI coding agents pick?
As of 2026 Q2, Clerk is the default auth pick of AI coding agents: the top pick in 51% of 243 recommendation probes (directional).
An auth library like NextAuth or Lucia is named in 62% of auth answers. DIY is the real runner-up.
Where do AI coding agents deploy?
As of 2026 Q2, Vercel is the default hosting pick of AI coding agents: the top pick in 49% of 243 recommendation probes (directional).
Self-hosting comes up in 19% of hosting answers.
Which voice-AI platform do AI coding agents pick?
As of 2026 Q2, Vapi is the default voice-AI pick of AI coding agents: the top pick in 44% of 243 recommendation probes (directional).
How to read this index.
Top-pick share
How often the agent makes this tool its primary pick when asked to build in the category. Being named isn't being picked; plenty of tools are mentioned constantly and chosen never.
Directional
Aggregates from controlled benchmark runs, reported with sample size. Below n=30 a cell is not reported at all.
Agents covered
Claude Code and Codex, across multiple wordings of each task. Per-model splits and full funnels live in the private scorecards.
Full method, detection, and gates: the methodology. The findings behind the index: the Agent Default Report.