Skip to main content

TapHerd

Your AI can now use Android apps.

TapHerd is a free, open-source connector that lets Codex, Claude, Cursor, or any MCP client control your Android phones and emulators. Your AI sees the screen, taps, types, opens apps, and copies media, on up to three devices for free.

# macOS and Linux
curl -LsSf https://raw.githubusercontent.com/GreenPeaas/tapherd/main/scripts/install.sh | sh
# Windows (PowerShell, no administrator rights needed)
irm https://raw.githubusercontent.com/GreenPeaas/tapherd/main/scripts/install.ps1 | iex

The installers set up uv and TapHerd for your user, offer to install Android's adb tool, and run tapherd doctor. They explain and ask before anything that needs sudo. What the installer does

macOS: tested · Windows and Linux: beta

Who it's for: developers and technical creators. You need a computer plus an Android emulator or an Android phone.

Animated replay: one open_app_many call opens Android Settings on Pixel 1 and Pixel 2, and each Pixel reports ok: true

Animated replay of a verified run: one TapHerd call opened Settings on two Pixel emulators. Watch one prompt post to three platforms (real screen recording, sped up, names blurred)

What it's for

  • Posting. Ask your AI to post a video to TikTok, Instagram Reels, and YouTube Shorts from your phones, with captions it writes. One prompt did this on 2026-11-19, on one emulator: Shorts set to Private, TikTok to Only you, and Reels public. The same tools work in other Android apps, such as X and the main YouTube app.
  • QA testing. Let an agent walk through your Android app on several emulators at once.
  • Store screenshots. Capture the same screens on several devices.
  • Mobile-only apps. Use apps that have no web or API version.
  • Your own analytics. Read numbers from your own accounts' in-app dashboards.

TapHerd only acts on devices you connect, through apps you're signed into. Its recipes and scheduler only support posting to accounts you own. Acceptable use

Install → first AI-controlled tap (about 5 minutes)

  1. Get an Android device. Start a free Pixel emulator from Android Studio, or connect a phone with USB debugging on. macOS · Windows · Linux

  2. Install TapHerd with the one-liner above, or manually with uv:

    uv tool install git+https://github.com/GreenPeaas/tapherd
    tapherd doctor
    
  3. Connect your AI client. Print its config and paste it where TapHerd says. Exact steps for each client

    tapherd config mcp --client claude-code     # or codex, claude-desktop, cursor
    
  4. Ask it to do something:

    Use TapHerd to list my devices. On the first one, open Settings, read the screen, tap Network & internet, and tell me what opened.

    You should see Network & internet open on the device.

Try it from the command line with Claude Code

This runs the same check with a temporary MCP config and only the four tools it needs:

TAPHERD_FIRST_TAP_DIR=$(mktemp -d)
tapherd config mcp --client claude-code > "$TAPHERD_FIRST_TAP_DIR/mcp.json"
(
  cd "$TAPHERD_FIRST_TAP_DIR" || exit
  claude --print \
    'Use only TapHerd MCP. List devices. On the first active device, open com.android.settings, use read_screen to find Network & internet, tap that row, and tell me what opened.' \
    --mcp-config "$TAPHERD_FIRST_TAP_DIR/mcp.json" --strict-mcp-config \
    --no-session-persistence --tools '' \
    --allowedTools mcp__tapherd__list_devices mcp__tapherd__open_app \
      mcp__tapherd__read_screen mcp__tapherd__tap
)

Menu labels can vary by Android version or language. If Settings resumes on a subpage, go back to its home screen and try again.

Not a developer? The fastest path

  1. Run the one-line installer above. It sets everything up and runs tapherd doctor.
  2. For a phone, either install Android Studio and start an emulator (the setup guide walks through it), or use a cloud phone that offers ADB access and connect it with tapherd connect <host:port>. Cloud phones haven't been tested with TapHerd yet, so provider guides are coming.
  3. Install Claude Desktop, run tapherd config mcp --client claude-desktop, and paste the result where it says.
  4. Ask Claude to do something on the phone.

Free and Pro

Free (this repo, MIT) Pro (coming soon)
Active devices Up to 3 50+ target
AI control of any Android app ✓ ✓
Posting recipes 1 unverified YouTube Shorts sample Maintained recipe library
Scheduling Run on demand Scheduled posting
Dashboard — Planned

tapherd pro opens the waitlist. The CLI shows a Pro notice at most once a day; set TAPHERD_NO_TEASE=1 to turn it off.

Docs

Get started Use it Project
Setup on macOS MCP tools Security and data flow
Setup on Windows (beta) Recipes Acceptable use
Setup on Linux (beta) Advanced MCP options Contributing
Connect your AI Architecture Changelog
Troubleshooting Demo clips License (MIT)

Privacy

The free connector runs entirely on your computer and has no telemetry. Screenshots and the optional action log stay local. TapHerd never asks for or stores passwords. You sign in to apps on the device yourself. Security and data flow


TapHerd is not affiliated with or endorsed by TikTok, Meta, Google, or X. All trademarks belong to their owners.

Metadata

Release files for tapherd 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for tapherd 0.1.0
File Size Uploaded Provenance Actions
tapherd-0.1.0.tar.gz 581.4 kB Without provenance Details for tapherd-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for tapherd 0.1.0
File Interpreter ABI Platform Actions
tapherd-0.1.0-py3-none-any.whl Python 3 none any Details for tapherd-0.1.0-py3-none-any.whl

Total release size: 822.1 kB

Release files / tapherd-0.1.0.tar.gz

Download URL tapherd-0.1.0.tar.gz
Size 581.4 kB
Tags Source
SHA-256 checksum
How to use checksums
0c51c9ddd484bc4c5660abd6a293e5d4d1adafdc3da13ae45e5c2796dd92e5cf
BLAKE2b-256 checksum
How to use checksums
a4d3f6b254389cfbd2d8f578252ef30132563863090a730c2e9cbabe39f1200e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.13.7 {"installer":{"name":"uv","version":"0.13.7","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Release files / tapherd-0.1.0-py3-none-any.whl

Download URL tapherd-0.1.0-py3-none-any.whl
Size 240.8 kB
Tags Python 3
SHA-256 checksum
How to use checksums
73412877cc52fc4a8f0edd6d4d53a0adfc78d6472323b8c16d1cc4bb5c3fad57
BLAKE2b-256 checksum
How to use checksums
0922ab099a7a48cc6125b1468e2219b50eeb94eca8c783887a92a8219cdefb26
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.13.7 {"installer":{"name":"uv","version":"0.13.7","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Security for tapherd 0.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page