PepperBLOCK
iOS dynamic library MCP for agents
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Pepper gives AI agents eyes and hands inside iOS Simulator apps.
It injects a shared library into any running simulator app — no source changes, no SDK, no build step. Your agent sees the screen as structured data, taps buttons, inspects live objects, intercepts network calls, reads the heap, and debugs layout issues. Dylib injection requires the simulator; device support uses a different mechanism.
https://github.com/user-attachments/assets/42ab3f1b-21f8-48e6-820f-7ca4012fb03b
Claude navigating and inspecting [Ice Cubes](https://github.com/Dimillian/IceCubesApp) (Mastodon client) with zero source access.
Works with Claude Code · Cursor · Claude Desktop · any MCP client
Quickstart
Requires macOS 14+, Python 3.10+, and an iOS Simulator runtime.
pip install pepper-ios pepper-ctl deploy # inject into the frontmost simulator app pepper-ctl look # see what's on screen
Then ask your agent:
"Tap through the onboarding flow and make sure every screen looks right"
That's it. The agent uses Pepper's MCP tools to see, tap, and inspect — no extra config needed.
Other MCP clients (Cursor, Claude Desktop, etc.)
{
"mcpServers": {
"pepper": {
"command": "pepper-mcp"
}
}
}Homebrew
brew install --HEAD skwallace36/pepper/pepper
Tap: `skwallace36/homebrew-pepper`
What It Does
$ pepper-ctl look Screen: navigation_stack (9 interactive, 4 text) seg "Trending" → tap text:"Trending" btn "Sheryl Weikal, Right wing tech bros: we love St..." → tap text:"Sheryl Weikal, Right wing tech bros: we ..." seg "Settings" → tap text:"Settings" $ pepper-ctl tap --text "Settings" Action: Tapped Settings Scree
85ea78e915deOBSERVED · 2026-10-07Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add pepper-ios -- uvx pepper-ios
{
"mcpServers": {
"pepper-ios": {
"command": "uvx",
"args": [
"pepper-ios"
]
}
}
}Exposed tools (4)
4 read · 0 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
nav_back | read | Go back one screen in a navigation stack. For modals/sheets use dismiss; for jumping to a screen use navigate. |
nav_dismiss | read | Close a modal, sheet, popover, or overlay. For navigation stack screens use back; for system dialogs use dialog. |
nav_keyboard | read | Dismiss the on-screen keyboard. Call after input_text when you need to tap elements the keyboard covers. |
nav_screen | read | Identify which screen is currently displayed. Returns screen name and view controller class — no element data. |
Trust audit
BLOCKgrade F · trust 52/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- UNDECLARED (10 observation(s))
- Network
- declared (7 observation(s))
- Shell
- declared (1 observation(s))
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (25)
@# Inject via SIMCTL_CHILD_ at exec() time instead (see simctl launch call below).
os.remove(port_file)
shutil.rmtree(dst)
os.remove(path)
os.remove(path)
os.remove(port_path)
@# DYLD_INSERT_LIBRARIES is a restricted var — launchctl setenv silently drops it.
@# Non-DYLD_ vars work fine via launchctl setenv.
@xcrun simctl spawn "$(SIMULATOR_ID)" launchctl setenv PEPPER_PORT "$(PORT)"
@xcrun simctl spawn "$(SIMULATOR_ID)" launchctl setenv PEPPER_SIM_UDID "$(SIMULATOR_ID)"
@xcrun simctl spawn "$(SIMULATOR_ID)" launchctl setenv PEPPER_ADAPTER "$(ADAPTER_TYPE)"
.swift-format
.swiftlint.yml
tests/unit/Sources/PepperCore/PepperCommand.swift
PORT ?= $(shell echo "$(SIMULATOR_ID)" | python3 -c "import sys,hashlib; uid=sys.stdin.read().strip(); print(8770 + int(hashlib.md5(uid.encode()).hexdigest()[:4],16) % 100 if uid else 8765)"
code_hash = hashlib.md5(source.encode()).hexdigest()[:8]
text_hash = hashlib.md5(text.encode()).hexdigest()
h = int(hashlib.md5(udid.encode()).hexdigest()[:4], 16)
print(8770 + int(hashlib.md5(uid.encode()).hexdigest()[:4], 16) % 100 if uid else 8765)
@ln -sf ../../scripts/hooks/post-merge-trigger.sh .git/hooks/post-merge
@ln -sf ../../scripts/hooks/pre-push-rebase.sh .git/hooks/pre-push
@ln -sf ../../scripts/hooks/post-checkout-guard.sh .git/hooks/post-checkout
SCRIPT_PATH="../../scripts/pre-commit"
PUSH_SCRIPT_PATH="../../scripts/hooks/pre-push-rebase.sh"
f.write(base64.b64decode(screenshot_b64))
Gates applied: no_behavioural_pass.
85ea78e915defull audit observations/trust-audit/mcp-server/skwallace36__pepper.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-07 | 85ea78e915de | BLOCK | F | 52 | first audit |
Questions
What is the Pepper MCP server?
iOS dynamic library MCP for agents
What tools does Pepper expose?
4 in total: 4 read-only, 0 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is Pepper safe to connect to an agent?
No — not without reading the findings first. The audit graded it F (52/100) and found 1 critical or high issue in the source. Each one is listed on this page with the file and line it is on.
What credentials does Pepper need?
No credential environment variables were found in its source, so it appears to need none.
How does Pepper run?
It speaks stdio, so it runs as a local process your client starts. It is published on PyPI as pepper-ios.
How current is this page?
The grade is for one exact copy of the source (85ea78e915de), read on 2026-10-07. The repository is watched and re-audited when it changes.