Gives your AI agent direct access to Android devices and emulators through ADB. You get 14 core tools for screen capture, UI inspection, touch input, text entry, and shell commands, plus an optional app_manager pack with 9 more for APK extraction, package management, and intent injection. Built for mobile QA automation, app testing, and security research where you need to tap, swipe, install, inspect permissions, or pull artifacts without modifying the target app. Runs in unrestricted mode with full ADB access or restricted mode with allowlists. Returns structured, minimal responses instead of raw XML dumps to keep context windows manageable during long automation sessions. Optional JSONL logging captures every tool call and security event.
Agent Droid Bridge is an MCP server that connects AI agents to Android devices and emulators over ADB. It is built for mobile automation, app testing, dynamic analysis, and reverse engineering: exposing the full surface of ADB as structured tools that any MCP-compatible AI client can call directly. If ADB can do it, an agent can do it.
Purpose-built tools return structured, minimal responses. No raw XML dumps, no wasted context — agents stay fast across long sessions.
The demo above runs through a few straightforward tasks to show what a connected agent can do, and this is just scratching the surface:
Tools
app_manager pack adds 9 tools for package management, app lifecycle, APK extraction, permissions, and intent injection. Load it with ADB_EXTRA_TOOL_PACKS=app_managerDevice handling
Security
unrestricted (full ADB access, optional denylist) and restricted (allowlist-only, blocks everything not explicitly permitted)ADB_ALLOW_SHELL=false to disable all shell commands regardless of modeADB_DENIED_TOOLSshlex. No shell injection possibleObservability
MCP_LOG_ENABLED=true and MCP_LOG_DIRMobile QA and test automation Automate UI flows across real devices and emulators without modifying the app or writing test code. Tap, swipe, type, read screen content, take screenshots — all from a natural language prompt.
App security research Extract APKs, inspect declared permissions, fire arbitrary intents, and observe runtime behavior on screen. No instrumentation, no jailbreak required.
Dynamic analysis Launch apps in controlled states, drive UI interactions, capture screen state at each step, and pull artifacts — all scriptable through an AI agent.
Development and debugging Install builds, verify UI states, check app info, and run ADB commands without leaving your coding environment.
uvx agent-droid-bridge
No cloning or virtual environments needed. Requires Python 3.11+ and ADB installed on your host.
uvx is provided by uv. If you don't have it: curl -LsSf https://astral.sh/uv/install.sh | sh
To install from source instead, see docs/setup.md — Option B.
To verify the install: uvx agent-droid-bridge --help
{
"mcpServers": {
"agent-droid-bridge": {
"command": "uvx",
"args": ["agent-droid-bridge"],
"env": {
"ADB_EXECUTION_MODE": "unrestricted",
"ADB_ALLOW_SHELL": "true",
"ADB_PATH": "adb",
"ADB_EXTRA_TOOL_PACKS": "",
"MCP_LOG_ENABLED": "false",
"MCP_LOG_DIR": "~/logs/agent-droid-bridge"
}
}
}
}
To enable session logging, set MCP_LOG_ENABLED to "true" and update MCP_LOG_DIR to a writable path on your machine.
| Variable | Default | Description |
|---|---|---|
ADB_EXECUTION_MODE | unrestricted | Security mode. unrestricted allows all shell commands (with optional denylist); restricted allows only commands in ADB_SHELL_ALLOWLIST. |
ADB_ALLOW_SHELL | true | Set to false to block all adb shell commands regardless of execution mode. |
ADB_PATH | adb | Path to the ADB binary. Replace with a full path if adb is not on your system PATH (e.g. C:\platform-tools\adb.exe on Windows). |
ADB_EXTRA_TOOL_PACKS | (empty) | Comma-separated list of extra tool packs to load. Set to app_manager to enable 9 additional app management tools. |
MCP_LOG_ENABLED | false | Set to true to enable session logging. Requires MCP_LOG_DIR. |
MCP_LOG_DIR | (none) | Directory where session logs are written. Required when MCP_LOG_ENABLED is true. |
agent-droid-bridge MCP toolsFull setup guide and environment variable reference: docs/setup.md
| Tool | What it does |
|---|---|
get_ui_hierarchy | Returns the current screen as an XML UI hierarchy |
take_screenshot | Captures the screen as a base64-encoded PNG |
tap_screen | Sends a tap gesture at pixel coordinates |
swipe_screen | Sends a swipe gesture between two points over a given duration |
type_text | Types text into the focused input field |
press_key | Sends an Android keycode event (Back, Home, Enter, etc.) |
launch_app | Launches an app by its package/activity component name |
execute_adb_command | Runs an arbitrary ADB or ADB shell command |
list_devices | Lists all Android devices currently visible to ADB with their serial, state, and model |
snapshot_ui | Takes a lightweight UI snapshot and returns a token for use with detect_ui_change |
detect_ui_change | Polls for a UI change after an action; accepts a snapshot token as baseline; returns hierarchy only when requested |
get_screen_elements | Parses the UI hierarchy and returns structured elements with coordinates and interaction properties; supports tappable, interactive, input, and all modes |
get_screen_text | Returns all visible text on screen sorted top-to-bottom, as plain text |
check_device_capabilities | Returns structured device information — identity, security posture, and hardware specs — in a single call; supports identity, security, hardware, and all modes |
Optional packs extend the core toolset. Enable them by setting ADB_EXTRA_TOOL_PACKS in your MCP client config. See docs/extra-tool-packs.md.
app_manager
Package management, app lifecycle control, APK extraction, permission management, and intent injection.
| Tool | What it does |
|---|---|
list_packages | Lists installed packages with optional filtering, search, and detail levels |
get_app_info | Returns full static metadata for a single installed app |
install_app | Installs an APK from a host path onto the device |
uninstall_app | Removes an installed app by package name |
pull_apk | Extracts the installed APK from the device to the host |
manage_permission | Grants, revokes, checks, or lists runtime permissions for an app |
launch_app_extra | Launches an app by package name, auto-resolving the launcher activity |
manage_app | Controls app runtime state — stop, clear data, clear cache, enable, disable |
inject_intent | Fires an intent at a component via am start, am broadcast, or am startservice |
Full parameter reference: docs/tools.md
Configure the server entirely from your MCP client's env block. No files to edit. The env block in the Quick Start above covers the most common settings. For the full reference including security filtering, tool visibility, and timeouts, see docs/configuration.md.
To use a YAML config file instead, set ADB_CONFIG_SOURCE=yaml. See docs/configuration.md for details.
Session recording is separate. Enable it with MCP_LOG_ENABLED=true and MCP_LOG_DIR. Full reference: docs/logging.md.
| File | Description |
|---|---|
| docs/setup.md | Prerequisites, installation, and MCP client configuration |
| docs/tools.md | Full parameter reference for all tools |
| docs/configuration.md | Reference for environment variables and adb_config.yaml |
| docs/logging.md | Session recorder — log files, levels, retention, and activation |
| docs/extra-tool-packs.md | Extra tool packs — enabling packs, the pack contract, and writing your own |
| docs/troubleshooting.md | Common setup issues and ADB problems |
| docs/workflows.md | Common multi-tool workflows with examples |
| CHANGELOG.md | Release history and version changes |
Contributions are welcome. See CONTRIBUTING.md for guidelines on setup, code standards, and submitting pull requests.
To report a security vulnerability, follow the process in SECURITY.md — do not open a public issue.
ADB_EXECUTION_MODEdefault: unrestrictedSecurity mode. 'unrestricted' allows all shell commands with an optional denylist. 'restricted' allows only commands in ADB_SHELL_ALLOWLIST.
ADB_ALLOW_SHELLdefault: trueSet to 'false' to block all adb shell commands regardless of execution mode.
ADB_PATHdefault: adbPath to the ADB binary. Use a full path if adb is not on your system PATH.
ADB_EXTRA_TOOL_PACKSComma-separated list of extra tool packs to load. Set to 'app_manager' to enable 9 additional app management tools.
ADB_DENIED_TOOLSComma-separated list of tool names to hide from the agent at startup.
ADB_SHELL_ALLOWLISTComma-separated list of shell commands permitted in restricted mode. Empty blocks all shell commands.
ADB_SHELL_DENYLISTComma-separated list of shell commands blocked in unrestricted mode.
MCP_LOG_ENABLEDdefault: falseSet to 'true' to enable session logging. Requires MCP_LOG_DIR.
MCP_LOG_DIRAbsolute path to the directory where session logs are written. Required when MCP_LOG_ENABLED is true.
ADB_LOG_LEVELdefault: INFOServer process log level.
ADB_COMMAND_TIMEOUTdefault: 30Maximum seconds to wait for any ADB command.
ADB_SCREENSHOT_TIMEOUTdefault: 60Maximum seconds to wait for a screenshot capture.
ADB_UI_CHANGE_TIMEOUTdefault: 10Default timeout in seconds for detect_ui_change.
ADB_UI_CHANGE_POLL_INTERVALdefault: 0.5How often in seconds detect_ui_change polls the UI hierarchy.
ADB_AAPT_TIMEOUTdefault: 10Maximum seconds to wait for aapt dump badging, used by the app_manager pack.
ADB_CONFIG_SOURCEdefault: envSet to 'yaml' to load configuration from adb_config.yaml instead of environment variables.
io.github.ericm1018/skillfm-llm-cost-optimizer-openai-anthropic-usage
io.github.mikerawsonnz/llm-orchestration-agent
io.github.mikerawsonnz/authenticated-llm-agent
labforgedev/copilot-memory-mcp
csoai-org/agent-prompt-injection-firewall-mcp
io.github.mikerawsonnz/authenticated-multi-llm-agent