appstore-review-cli

Turn App Store and Google Play reviews into product intelligence - from the terminal or through your AI coding agent.

App store reviews are the largest public dataset of unfiltered user feedback. But reading them on the store websites is painful: no filtering, no export, no way to search across versions or countries. This tool fixes that.

Why Use This?

• Two stores: Apple App Store and Google Play - same filters, same output, one tool.
• Competitor research: Pull 1-star reviews for any app and find the feature gaps your product can fill.
• Bug triage: Filter reviews by keywords like "crash", "freeze", "login" and group by app version.
• Version monitoring: Compare sentiment between releases - see what got better, what got worse, and what's new.
• Multi-country insights: Same app, different markets - compare complaints across us, de, jp, etc.
• AI-native: Your coding agent (Copilot, Claude Code, Cursor) can fetch and analyze reviews in natural language. No Ollama needed - the agent IS the LLM.

No API keys. No accounts. No servers. Just pip install and go.

Install

pip install appstore-review-cli

For Google Play support:

pip install "appstore-review-cli[google]"

Install as an agent skill:

appstore-reviews setup copilot   # GitHub Copilot skill
appstore-reviews setup claude    # Claude Code skill

Installs a SKILL.md that the agent auto-discovers when relevant. See Agent Integration for Cursor, Windsurf, and global install options.

Quick Start

# Find an app
appstore-reviews search "Slack"

# Get negative reviews from the last 30 days
appstore-reviews reviews 803453959 --stars 2 --days 30

# Filter by keywords
appstore-reviews reviews 803453959 --keywords crash,freeze --stars 2

# Sort by most helpful
appstore-reviews reviews 803453959 --stars 2 --sort votes

# Get only 3-star reviews (the nuanced ones)
appstore-reviews reviews 803453959 --min-stars 3 --stars 3

# Export to CSV or JSON
appstore-reviews reviews 803453959 --stars 2 --format csv
appstore-reviews reviews 803453959 --stars 2 --format json

# Compare two apps side by side
appstore-reviews compare 803453959 310633997 --stars 2 --pages 3

# Google Play - use --store google with package names
appstore-reviews --store google search "Slack"
appstore-reviews --store google reviews com.Slack --stars 2 --days 30
appstore-reviews --store google compare com.Slack com.microsoft.teams --stars 2

# Compare sentiment between app versions
appstore-reviews version-diff 803453959 --pages 5
appstore-reviews version-diff 803453959 --old 4.23.0 --new 4.29.149

# Show rating trend over time (weekly or monthly)
appstore-reviews trend 803453959 --pages 5
appstore-reviews trend 803453959 --period month --stars 2

# Export compare/version-diff/trend to JSON or CSV
appstore-reviews compare 803453959 310633997 --format json
appstore-reviews version-diff 803453959 --format csv
appstore-reviews trend 803453959 --format csv

Agent Integration

One command to teach your AI coding agent every command, filter, and workflow:

# Project-level (per-repo)
appstore-reviews setup copilot    # .github/skills/appstore-reviews/SKILL.md
appstore-reviews setup claude     # .claude/skills/appstore-reviews/SKILL.md
appstore-reviews setup cursor     # .cursor/rules/appstore-reviews.md
appstore-reviews setup windsurf   # .windsurfrules

# Personal (shared across all projects)
appstore-reviews setup copilot --global   # ~/.copilot/skills/appstore-reviews/SKILL.md
appstore-reviews setup claude --global    # ~/.claude/skills/appstore-reviews/SKILL.md

Then just ask in natural language:

• "What are the top complaints about Slack this month?"
• "Find crash reports for WhatsApp in the last 30 days"
• "Compare Notion vs Obsidian - what do users hate about each?"
• "What features are German Duolingo users requesting?"

The agent runs the CLI, fetches reviews, and analyzes them directly. No Ollama, no extra setup.

Use --force to overwrite an existing file, --append to add to one.

How it works per agent

GitHub Copilot and Claude Code support agent skills - a directory-based system where each skill has a SKILL.md file with YAML frontmatter (name and description). The agent reads the description and only loads the skill into context when it's relevant to the current task. Both support project-level and personal (global) skill directories.

Cursor and Windsurf do not have a skills system yet. Instead:

• Cursor uses rules (.cursor/rules/) that are loaded based on glob patterns or always active. We install a single rules file there.
• Windsurf uses a single flat rules file (.windsurfrules) in the project root, loaded on every interaction.

For Cursor and Windsurf, --global is not supported since they have no personal/global rules directory.

Without an agent

Use the built-in Ollama analysis, or pipe to any LLM:

# Ollama (local, private)
ollama pull qwen3.5:4b
appstore-reviews analyze 803453959 --stars 2 --mode summary
appstore-reviews analyze 803453959 --stars 2 --mode gaps
appstore-reviews analyze 803453959 --stars 2 --mode bugs --keywords crash,freeze

# Or pipe raw output to any tool
appstore-reviews reviews 803453959 --stars 2 --format text | your-llm "Summarize:"

All Options

Global flags (apply to all commands)

search - Find an app by name

For Google Play, app IDs are package names (e.g. com.Slack). For Apple, they're numeric (e.g. 803453959).

reviews <APP_ID> - Fetch and filter reviews

All filters stack with AND logic.

analyze <APP_ID> - LLM analysis via Ollama

Plus all the same filters as reviews (--stars, --min-stars, --days, --keywords, --version, --pages, --sort, --stats, --country).

compare <APP_ID> <APP_ID> [...] - Compare multiple apps

Outputs: overview table, per-app rating distribution, top complaint categories, top keywords, shared vs unique complaints.

version-diff <APP_ID> - Compare sentiment between versions

Outputs: version comparison table, rating distributions, complaint category changes (with arrows), new/resolved issues, top keywords per version. Versions are auto-detected from the two most reviewed if not specified.

trend <APP_ID> - Show rating trend over time

Outputs: per-period table with average rating, review count, trend arrows (▲/▼), ASCII sparkline bars, mini star distributions, and overall trend summary.

setup <agent> - Install agent skill / instructions

Country codes: us (default), gb, de, fr, jp, au, ca, nl, br, kr

Python API

from appinsight import get_reviews, get_reviews_df, search

# Search
apps = search("Slack", limit=3)

# As dicts (no pandas needed)
reviews = get_reviews(618783545, stars=2, days=30)

# As pandas DataFrame
df = get_reviews_df(618783545, stars=2, pages=5)
df.groupby("version")["rating"].mean()
df[df["content"].str.contains("crash", case=False)]
# Google Play
reviews = get_reviews("com.Slack", stars=2, days=30, store="google")
df = get_reviews_df("com.Slack", stars=2, pages=5, store="google")

Install with pandas: pip install appstore-review-cli[pandas] Install with Google Play: pip install appstore-review-cli[google]

Good to Know

• Pipe-safe: Data goes to stdout, progress to stderr.
• Review limit: Apple returns max ~500 reviews per country (10 pages × 50). This is Apple's limit.
• Deduplication: Reviews are automatically deduplicated across pages.
• Validation: --stars/--min-stars accept 1-5, --pages accepts 1-10. Invalid values are rejected.
• Google Play search: The first search result sometimes lacks a package name (library limitation). Use the package name directly if your app doesn't appear (find it in the Google Play URL).
• No results? Filters are too narrow - try fewer keywords, more days, or a higher star ceiling.

Development

git clone https://github.com/vkayatas/appstore-review-cli.git
cd appstore-review-cli
uv sync
uv run pytest