MIT open source, paid data

Open source review monitoring: looot-reviews

Read public reviews of your company and your competitors and group them into themes.

The code is MIT open source. The data is not free and not self-hosted: every real run calls looot and pays per call at the catalog price shown before the run.

View the repository on GitHub

Short answer

What is looot-reviews?

looot-reviews is an MIT-licensed web app and CLI. Name your company or place and its competitors. It pulls public reviews, groups them into themes with counts and example quotes, tracks the average rating by month and lists themes where a competitor's customers are unhappy and yours are not. The code is open source. The data is paid per looot call.

What it does

What looot-reviews does and how to run it

Name a target and its competitors on the command line or in a YAML file. Google Maps needs only a name. App Store, Google Play and Tripadvisor need an id for each place or app.

It groups reviews into themes by counting words, with no LLM needed, and shows counts and example quotes. It tracks the average rating by month and lists themes where a competitor is weak and you are not. An optional LLM step can merge themes if you set your own key.

The CLI prints the plan and a maximum cost before it runs, and the maximum counts every page it may fetch. Results go to JSON, CSV and a markdown brief.

This is the job that products such as ReviewTrackers do.

Commands for looot-reviews
git clone https://github.com/loootai/looot-reviews && cd looot-reviews
npm install --ignore-scripts
npm run demo
npm run reviews -- --demo --out out
npm run reviews -- --config reviews.example.yaml --dry-run
export LOOOT_TOKEN=...
npm run reviews -- --target "Blue Door Cafe @ Austin TX" --competitors "Copper Pot @ Austin TX; Tin Roof @ Austin TX" --max-spend 0.10 --out out

Facts

Facts from the repository

Read from GitHub on 2026-10-11.

Facts about looot-reviews
JobRead public reviews of your company and your competitors and group them into themes.
LicenseMIT
LanguageTypeScript
A real run needs
  • Node.js 22.12 or newer
  • A looot token with the scopes runs.execute, catalog.read, runs.read and usage.read, for real pulls only
  • A funded looot balance, for real pulls only
  • An admin password of 12 characters or more for the web app outside the demo
Demo modeYes. npm run demo shows invented shops with generated reviews, and --demo writes the same data as files. Both need no token and make no looot call.
Repositorygithub.com/loootai/looot-reviews
Facts read on2026-10-11

Data and price

Endpoints it calls, with the live price

Every real run calls these looot endpoints. The price column is read live from the catalog, never typed into this page.

Endpoints called by looot-reviews with their live catalog price
EndpointUsed forLive price
serper-placesserperGoogle Maps place lookupreading live price
serper-reviewsserperGoogle Maps reviews, per pagereading live price
searchapi-io-apple-product-reviewssearchapi-ioApp Store reviews, per pagereading live price
anysite-api-google-play-apps-reviewsanysiteGoogle Play reviewsreading live price
searchapi-io-tripadvisor-reviewssearchapi-ioTripadvisor reviews, per pagereading live price

Prices are read from the looot public catalog when the page loads. Some endpoints are priced per result, so a run that returns many rows costs more than one call. The CLI dry run prints the exact quote for your input before anything is charged.

Limits

What looot-reviews does not do

  • It reads public reviews as providers return them. It does not reply, post or monitor continuously. Run the CLI on a schedule if you want that.
  • It keeps rating, text, title and date, and does not keep reviewer names.
  • It fetches the first pages the provider gives, up to 5 pages per source, not every review of a large place.
  • Trustpilot, G2 and Yelp are missing because the catalog has no endpoint for them.
  • Sentiment is the star rating, not text analysis. Themes count words, work on English text only, and a set under about 30 reviews gets a warning.
  • The web form reads Google Maps only. Use the CLI with a YAML file to add the other sources.
  • Response parsing was written from provider documentation and fixtures. The README says it was not checked against a recorded paid response.

With an agent

Run it with your agent

Copy the prompt, give it to the agent you already use, and it dry-runs first.

Prompt: looot-reviews
Use looot to find what customers complain about at my competitors. Read https://looot.ai/skill.md first. Then clone https://github.com/loootai/looot-reviews, write my company and my competitors into a YAML file, run its dry run and show me the quote and the cap. Run the real pull only after I say yes, and list the themes where competitors are weak and I am not.

FAQ

Questions about looot-reviews

It pulls public reviews of your company and your competitors through looot, groups them into themes with counts and example quotes, tracks the average rating by month and shows where a competitor's customers are unhappy.

The code is free under the MIT license. The data is not. Every real pull is a set of paid looot calls at the catalog price, quoted before the run. Demo mode and the dry run cost nothing.

Google Maps, the App Store, Google Play and Tripadvisor. Trustpilot, G2 and Yelp are not included, because the catalog has no endpoint for them.

No. Themes come from counting words. An optional LLM step can merge themes if you set your own key and model, and it is off by default.

One place lookup and a number of pages per source. The table on this page shows the live catalog price of each endpoint, and the dry run prints the plan with a maximum, because later pages are fetched only when the provider says there are more.

It does not reply to reviews, post, or monitor on its own. Sentiment is the star rating, themes work on English text, and it keeps no reviewer names.

Yes. npm run demo and the CLI --demo flag use invented shops and need no token. A real pull needs a token and a funded balance.

Give your agent the prompt on this page and the setup link https://looot.ai/skill.md. The repo ships a skill that dry-runs first, shows the total and runs only after you agree.

Keep going

Related on looot.ai

For agents
repo
https://github.com/loootai/looot-reviews
job
Read public reviews of your company and your competitors and group them into themes.
license
MIT
language
TypeScript
endpoint
serper: serper-places (maps_local_search)
endpoint
serper: serper-reviews (maps_place_reviews)
endpoint
searchapi-io: searchapi-io-apple-product-reviews (reviews_appstore_reviews)
endpoint
anysite: anysite-api-google-play-apps-reviews (api_google_play_apps_reviews)
endpoint
searchapi-io: searchapi-io-tripadvisor-reviews (search_tripadvisor-reviews)
price
per call, read live from the catalog on the web page
framing
MIT code, paid data, every real run calls looot
setup
https://looot.ai/skill.md
llms.txt
https://looot.ai/llms.txt
read
2026-10-11
markdown /open-source-alternatives/looot-reviews.md

Run looot-reviews on your own looot account.

Try the demo first, no account needed. For a real run, start your workspace, top up, and use the quote the repo prints before it spends anything.

Start your workspace See all open source repos