---
title: "From init to deploy: building with agents and Stripe Projects"
description: "The post describes an internal agentic transcription app built to reduce the manual effort required to correct auto-generated subtitles for Stripe Developers videos. The app uses a Next.js interface with transcript, chat, and corrected-output panels, while Stripe Projects CLI links Vercel and OpenRouter and populates required environment variables. An LLM client supports selectable models, iterative instructions, and a transcription agent whose system prompt emphasizes preserving transcript structure; the agent can also use Stripe documentation search through MCP. In a benchmark on a 771-line meetup script, Gemini 2.5 Pro reached 99% accuracy, while the tested models reduced editing from more than 30 minutes to just over two minutes of generated edits plus a human check. The reported overall reduction was 4x, with testing costs totaling $2.54."
---

# From init to deploy: building with agents and Stripe Projects

[Stripe](https://yomu.fyi/company/stripe) · Anna Spysz · Apr 29, 2026

**Type:** Tutorial

## Summary

The post describes an internal agentic transcription app built to reduce the manual effort required to correct auto-generated subtitles for Stripe Developers videos. The app uses a Next.js interface with transcript, chat, and corrected-output panels, while Stripe Projects CLI links Vercel and OpenRouter and populates required environment variables. An LLM client supports selectable models, iterative instructions, and a transcription agent whose system prompt emphasizes preserving transcript structure; the agent can also use Stripe documentation search through MCP. In a benchmark on a 771-line meetup script, Gemini 2.5 Pro reached 99% accuracy, while the tested models reduced editing from more than 30 minutes to just over two minutes of generated edits plus a human check. The reported overall reduction was 4x, with testing costs totaling $2.54.

## Context

Auto-captioning tools make mistakes with proper nouns and technical context, so subtitles require human review before release. Editing auto-generated subtitles takes 1.5 to 2 times the video's length in minutes; a 20-minute video therefore requires roughly 30 to 40 minutes of editing, consuming developer time on repetitive work.

## Approach / What changed

The author built a hosted Next.js transcription-fixing app with a split-pane UI, session-only state, iterative API routes, and an agent that applies user instructions to rough transcripts. Stripe Projects CLI links Vercel and OpenRouter, manages integrations and environment variables, and supports deployment. The app lets users select LLM models, provide structured fields and chat instructions, and uses a system prompt, examples, and Stripe documentation search through MCP to improve corrections.

## Takeaways

- Gemini 2.5 Pro achieved 99% accuracy on the reported 771-line benchmark, compared with 85% for Gemini 3 Flash Preview, 90% for Claude Sonnet 4.5, and 95% for Claude Opus 4.5 thinking.
- The tested workflow reduced a 20-minute talk from more than 30 minutes of editing to a little over two minutes of generated edits, followed by an estimated two to three minutes of human checking.
- The project used Stripe Projects to link Vercel and OpenRouter, populate local environment variables, deploy to Vercel, and provide commands for teammates to pull or rotate provider credentials.

**Tags:** [Deployment](https://yomu.fyi/topic/deployment), [Developer Experience](https://yomu.fyi/topic/developer-experience), [LLMs](https://yomu.fyi/topic/llm), [Open Source](https://yomu.fyi/topic/open-source)

- Source: [Stripe](https://stripe.dev/blog/building-with-agents-stripe-projects)
- Source URL: https://stripe.dev/blog/building-with-agents-stripe-projects
- Ingested by Yomu: 2026-08-28T20:00:51.774Z

[Read original post](https://stripe.dev/blog/building-with-agents-stripe-projects)