← Guides · 日本語 · Português · 中文

Updated 2026-11-19

On this page
  1. What it is
  2. Build one
  3. Update loop
  4. How they run (the parts that matter)
  5. The three proven uses
  6. Verify it
  7. Pitfalls
  8. Knowledge check
  9. Sources

Guide 10 — Devin Apps: ship live tools from a chat

Goal: describe an internal tool, dashboard, or staging copy — Devin builds it and serves it at a stable URL your team opens by signing in. Level: start here (platform guide) · Time: ~10 minutes to first App Source: Introducing Devin Apps (Nov 13, 2026)

What it is

A Devin App is a live, private web application hosted directly on Devin's cloud machine. You describe the tool, Devin builds it, your team opens it from a link — no hosting, no deploy pipeline. Because it runs on Devin's machine, it can reach your MCPs, integrations, codebase, and its own database (SQLite/Postgres on the same box), and call the Devin API itself.

Build one

  1. Sidebar → Apps → describe what the App should do (or pick a prompt template: metrics dashboard, project status, review queue) → Build with Devin
  2. Devin asks clarifying questions in the session, builds, and puts it online — the Apps page shows its URL, status Building until v1 is live
  3. Share — copy the link; scope it to named people, your org, or your enterprise. Apps are never public; opening one requires a Devin sign-in

You can also say "publish this as an App" inside any session that produced a browser preview.

Update loop

How they run (the parts that matter)

The three proven uses

Use Example
Internal tools Admin panels, request forms, approval queues — the shared-spreadsheet killers
Dashboards Metrics/support/sales boards pulling from your own data sources, refreshed on schedule
Beta/staging copies Your real stack (frontend+backend+DB+workers) running on Devin's machine — teammates open a staging copy from a link

Verify it

Pitfalls

Knowledge check

  1. What's the difference between a Devin App and a PR preview? (The App is served live from Devin's machine at a stable URL — stateful, persistent, sign-in gated — not a build artifact.)
  2. Why does 'your team opens it by signing in' matter for internal tools? (Auth is inherited from the org — no separate login system to build or secure.)
  3. Which class of internal tool fits Apps best? (Dashboards and staging surfaces where 'good enough now' beats 'perfect later' — the maintenance cost is near zero.)

Sources