Próxima Estación
Practical Japanese on Instagram, produced by a team of Claude agents orchestrated by n8n.
- Type
- Automation
- Year
- Categories

About the project
Próxima Estación is an Instagram account of practical Japanese for Spanish speakers who travel to Japan or live there: the phrases you actually hear at the konbini, the restaurant or the hotel. Its look is a metro map: every topic is a coloured line and every post a station.
Goal. To publish useful, verified content every day without a team of people behind it. A team of AI agents produces it and a person approves it: the AI does the work and the final decision stays human.
What it does. The team is sixteen professionals, each one a Claude agent with its own role, orchestrated by n8n:
- Planning, every Sunday (6 agents): strategy, data analysis, trend hunting, monetisation and the week's calendar, with a fixed format for each day.
- Every post (8 agents): research, fact-checking, a native linguist, copywriting, social SEO, scriptwriting, an art director who reviews the rendered images and an editor-in-chief.
- Approval: no post goes out until I approve it on Telegram; then it is published on time through the Instagram API and its metrics are collected for the next planning.
- After publishing (2 agents): a community manager who answers the simple comments, and the PDF guide given away by email.
- Images and video: posts (1080×1350), stories and the PDF come from templates rendered with Playwright, and the weekly video, with a voice-over, is put together with Remotion.

SEO. Social SEO: an agent writes the keywords and at most five hashtags per post.
Accessibility. Every image carries its alternative text, and the Japanese typeface is bundled with the templates so it reads the same on any phone.
Security and privacy. n8n is only reachable through an SSH tunnel, the Telegram bot opens no ports, images go through a CDN with a key bound to their folder and expire after 48 hours, and backups go to Google Drive.
Architecture and quality. Ten n8n workflows and two TypeScript services with Fastify (the studio, with the agents and the bot, and the renderer), on PostgreSQL and Docker, deployed on every change.
Status. Built and deployed, in its launch phase.



