Chrome extensions that automate web apps
Custom Chrome extensions and browser automation that click, fill, select and export inside the web apps your team already uses, even when the vendor offers no API.
9eye Studio · custom automation
Workflow automation services for the software your team already uses.
9eye Studio is the team behind the 9eye broadcast platform. We automate the work your team does by hand in software, or check it before it goes out: Chrome extensions for web apps without an API, AI agents and MCP servers, validators and integrations for spreadsheets, file exports and older systems. Fixed-price projects, for teams anywhere in the world.
Automation services, from the browser to AI agents.
Six kinds of work, often in one project: an extension that does the job, a validator that checks it, and an agent that runs it when something happens.
Validators and data quality checks
Every file, export, playlist or record is checked against your rules before it goes out, with a report that shows what is wrong and where.
AI agent development and MCP servers
AI agents that do one job well with Claude, ChatGPT or a local model, and MCP servers that connect them to your systems, with a person approving what matters.
Integrations between systems
Data your team types twice today moves on its own: from sales to operations, from a web app to your database, from spreadsheets into the tool that needs it.
Google Sheets, Excel and data pipelines
Spreadsheet automation, scraping and scheduled pipelines that collect, clean and join data from APIs, websites and archives into files or a database.
Private, local AI
Speech-to-text, translation and classification on your own hardware. No per-minute cloud bills, and your content never leaves your network.
Selected work.
Real projects, described without client names. Most of our work is under NDA, so details are changed and no client data is shown.
9eye: broadcast scheduling with AI agents
A multi-tenant scheduling and traffic platform for TV channels, from catalog to as-run, with ten AI agents, a check-and-fix loop of validators and an MCP connection for any AI.
See 9eyeBroadcast playlist checker for a multi-channel TV network
The problem. Each day’s playlist is exported from a cloud scheduling system and sent to playout. Mistakes were caught on air, or by people reading hundreds of lines.
What we built. A browser extension that catches every export before it is saved and runs more than 20 checks on it: booked spots against the sales sheet, spot lengths, idents never back to back, end credits and “next” bumpers that match the coming show, episode order against the schedule, expired promos and early starts. It times the day, inserts missing clips from per-series rules and writes a corrected file, with a report that points to each problem.
Read the case studyFrom booked campaign to placed spots
The problem. Ad campaigns were booked in one system, and the traffic team typed every spot again in the scheduling system, break by break.
What we built. A pipeline that reads each campaign per channel, maps channels, ad boxes and spot lengths, and writes the order with every spot placed in its exact break. It stays within the vendor’s session limit and tests each step against the live API.
An agent that rebuilds airing patterns
The problem. Turning a programme’s week on air into a reusable pattern meant reading the schedule and rebuilding it by hand.
What we built. An agent skill that reads how a programme airs across the week, first runs and repeats, in a scheduling web app, and recreates it as a pattern step by step for a person to check.
Offline video translation pipeline
The problem. Translating shows for subtitles and dubbing meant sending content to cloud services and paying by the minute.
What we built. Video in, a translated script and subtitles out: speech recognition with exact timestamps, speaker detection, then translation by a local language model that keeps real names. It runs fully offline on one 6 GB graphics card.
Market research from what users complain about
The problem. A product team wanted to know what people actually ask for, without reading thousands of reviews.
What we built. A collector that gathers low-star reviews and feature requests from app stores, Product Hunt, Reddit archives, Stack Exchange and GitHub issues, and sorts them by votes into a sheet the team can act on.
Market data pipeline with walk-forward testing
The problem. Backtests looked good and failed live, because models had seen the future.
What we built. A pipeline that collects and cleans exchange market data, then tests models strictly walk-forward: trained only on the past, judged only on what came after, so results are not flattered by hindsight.
Timing checker
Checks that each programme block’s time-code in and out add up, and flags blocks that run long or short.
Catalog helper
Automates repetitive list selection in a catalog web app, so a long job is one click.
One-click capture to database
A button inside a booking web app that pulls the campaign data from an API and saves it to PostgreSQL.
If people can do it in the software, it can be automated.
We start from the screen your team works in, not from the vendor’s roadmap. If there is an API, we use it. If there is not, we work with the app the way a person does, only faster and the same way every time.
- Web apps, with or without an API
- Chrome extensions (Manifest V3)
- Excel, Google Sheets, Drive, Dropbox and FTP
- XML, CSV and JSON exports and imports
- REST APIs, webhooks and PostgreSQL
- Python, Node.js and TypeScript
- MCP servers for Claude, ChatGPT and other AI
- Local AI: speech-to-text, speaker detection, local language models
Automation for broadcasters and media companies.
Our deepest work is in TV: playlist QC before playout, ad traffic that places itself, airing patterns rebuilt by an agent, and translation for subtitles on your own hardware. If your channel runs on a scheduling, traffic or playout system, we already speak its language.
See the broadcast Playlist Checker- Broadcast playlist QC and validation before playout
- Campaign-to-traffic bridges and ad placement
- Scheduling patterns, catalog and metadata automation
- Subtitles and translation without cloud costs
- Built on what we learned making 9eye, our scheduling platform
How a project works.
Fixed scope, fixed price, and something working early.
- 01
Free call
30 minutes on the job you want gone: what the team does, in which software, and how often.
- 02
Fixed quote
A written plan with scope, price and delivery date. No hourly billing.
- 03
Build
Short cycles, each ending in a working demo on your real screens and data.
- 04
Check
Tested on real cases, with validators where a mistake is expensive.
- 05
Hand over
Source code, a short guide and a video. You own what you paid for.
- 06
Support
An optional monthly plan for changes when your software changes.
Tell us the job your team repeats.
Which software, what people do in it, and how often. We reply with questions or a time for a call, then send a plan and a fixed price.
Your message could not be sent. Please write to [email protected] or on WhatsApp at +1 971 480 9697.
Thank you. Your project is in.
We reply to your e-mail address with questions or a time for a call.
Studio questions
What do your workflow automation services include?
Finding the manual work worth automating, building the automation (a Chrome extension, a script, an integration, a validator or an AI agent), testing it on your real data, and handing over the code with a guide. Each project is fixed in scope and price before we start.
Can you automate a web app that has no API?
Yes. If the software has an API we use it; if not, we automate the screen itself with a custom Chrome extension or browser automation: clicking, filling forms, selecting and exporting the way a person does, only faster and the same way every time. Data entry from Excel or Google Sheets into a web form is a common first project.
Do you build custom Chrome extensions for internal tools?
Yes. Manifest V3 extensions for your team’s own use: a button inside the web app you already work in, a check before an export is saved, or a form filled from a spreadsheet. They can be installed privately, without the Chrome Web Store.
What is an MCP server, and do we need one?
MCP (Model Context Protocol) is an open standard that lets AI assistants such as Claude or ChatGPT use your systems as tools: read a record, run a search, create an order. An MCP server is the small program that offers those tools. You need one when you want an AI agent to work with your own data and software, with access limited to what you allow.
Do you build AI agents?
Yes, where they help: agents that do one job, with a person approving the result, built on Claude, ChatGPT or a model on your own hardware. Where plain rules do the job better, we use plain rules. 9eye, our broadcast platform, is built the same way.
How much does an automation project cost?
Every project has a fixed price, agreed before we start and based on the software, the steps and the checks involved. The first call is free, and the written quote states the scope, the price and the delivery date. An optional monthly plan covers changes after hand-over.
Do you need access to our systems?
Only what the job needs, ideally a test account. We can start from screen recordings and sample files, and we sign an NDA before we see your data.
Who owns the code?
You do, once the project is paid. You get the source code, a short guide and a hand-over video.
Do you work with clients in other countries?
Yes. We work remotely with teams anywhere in the world, in English, with written updates and demos at times that suit your time zone.