mac-mcpBLOCK
Talk to your AI agent and let it work on your Mac. Open source MCP control for browser, files, apps, shell, and agents.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Mac MCP is a local macOS control server for AI agents. It exposes your Mac through a native MCP endpoint and a REST/OpenAPI surface, with shell, files, browser automation, macOS UI control, delegated OpenCode/Codex agents, memory, Agent Skills, voice interaction, self-update tooling, and a local operations dashboard. The native MCP endpoint is the full capability surface; REST/OpenAPI intentionally publishes a selected compatibility subset, so some capabilities remain MCP-only.
Talk to ChatGPT. Let it work on your Mac.
Mac MCP is model-agnostic and works with MCP-compatible AI clients. But ChatGPT is an especially natural way to use it.
ChatGPT Live can use plugins during voice conversations. With Mac MCP connected, the same ChatGPT conversation you already use can become a control surface for your actual Mac. You can talk naturally while Mac MCP handles the execution: browsing in Safari or Chrome, working with files, opening apps, running commands, and coordinating agents.
Instead of moving every task into a separate coding-agent session, you can keep the interaction in ChatGPT and simply talk. From your phone, you can ask ChatGPT to work on a reachable Mac somewhere else. From your desktop, you can keep talking while Mac MCP works in the background without taking over the computer.
For example:
- "Check my Reddit replies, answer the important ones, and close the browser when you are done."
- "Go through the repo, run the tests, and tell me what is broken."
- "Find three hotels for next weekend, compare the reviews, and save the shortlist on my Mac."
ChatGPT is the conversation. Mac MCP is the execution layer.
This is a powerful combination, not a Mac MCP-only voice feature. ChatGPT provides the natural voice interface and reasoning experience; Mac MCP gives it a local execution layer on macOS. Existing ChatGPT plugin permissions, approval
b6aa4f9a8257OBSERVED · 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. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add mac-mcp --env CLOUDFLARE_TUNNEL_TOKEN_FILE=${CLOUDFLARE_TUNNEL_TOKEN_FILE} --env GROQ_API_KEY=${GROQ_API_KEY} --env MAC_MCP_CHROME_BRIDGE_TOKEN=${MAC_MCP_CHROME_BRIDGE_TOKEN} --env MAC_MCP_CHROME_BRIDGE_TOKEN_FILE=${MAC_MCP_CHROME_BRIDGE_TOKEN_FILE} -- uvx mac-mcp{
"mcpServers": {
"mac-mcp": {
"command": "uvx",
"args": [
"mac-mcp"
],
"env": {
"CLOUDFLARE_TUNNEL_TOKEN_FILE": "${CLOUDFLARE_TUNNEL_TOKEN_FILE}",
"GROQ_API_KEY": "${GROQ_API_KEY}",
"MAC_MCP_CHROME_BRIDGE_TOKEN": "${MAC_MCP_CHROME_BRIDGE_TOKEN}",
"MAC_MCP_CHROME_BRIDGE_TOKEN_FILE": "${MAC_MCP_CHROME_BRIDGE_TOKEN_FILE}"
}
}
}
}Exposed tools (50)
36 read · 13 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
agent_action | read | nonlocal body_calls |
agent_catalog | read | return await asyncio.to_thread( |
ask_choice | read | return { |
browser_act | read | return await asyncio.to_thread( |
browser_click_selector | read | return _log(audit_logger, |
browser_do | read | nonlocal executed |
browser_execute_js | write | return _log(audit_logger, |
browser_find | read | return { |
browser_get_html | read | return _log(audit_logger, |
browser_list_tabs | read | return _log(audit_logger, |
browser_observe | read | executed.append( |
browser_press_key | read | nonlocal pasted |
browser_type_selector | read | return _log(audit_logger, |
browser_wait_for_selector | read | return _log(audit_logger, |
clipboard_get | read | return _log(audit_logger, |
clipboard_set | write | return _log(audit_logger, |
computer_plan | read | return await execute_computer_plan(mcp.call_tool, steps=steps, max_seconds=max_seconds) |
copy_file | read | return _log(audit_logger, |
delete_path | destructive | return _log(audit_logger, |
directory_tree | read | return _log(audit_logger, |
edit_file | write | return _log(audit_logger, |
get_agent | read | return _log(audit_logger, |
get_file_info | read | return _log(audit_logger, |
get_job_output | read | return _log(audit_logger, |
get_job_status | read | return _log(audit_logger, |
get_system_info | read | return _log(audit_logger, |
get_volume | read | return _log(audit_logger, |
http_request | read | return _log(audit_logger, |
lesson_record | read | pair = current_security_session() |
list_agents | read | return _log(audit_logger, |
list_directory | read | return _log(audit_logger, |
list_jobs | read | return _log(audit_logger, |
mac_act | read | executed.append( |
move_file | write | return _log(audit_logger, |
open_app | read | return _log(audit_logger, |
open_url | read | return _log(audit_logger, |
process_list | read | return _log(audit_logger, |
read_file | read | return _log(audit_logger, |
read_multiple_files | read | return _log(audit_logger, |
run_applescript | write | return _log(audit_logger, |
run_command | write | return { |
send_notification | write | return _log(audit_logger, |
set_reminder | write | return _log(audit_logger, |
skill_update_index | write | return await asyncio.to_thread(_log, audit_logger, |
spawn_agent | read | return { |
stop_job | write | return _log(audit_logger, |
tool_discover | read | q = str(query or |
tool_invoke | write | return await _invoke_registered_tool(mcp, tool_name, arguments) |
write_file | write | return _log(audit_logger, |
write_files_batch | write | return _log(audit_logger, |
Trust audit
BLOCKgrade F · trust 49/100 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 | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (16 observation(s))
- Network
- declared (5 observation(s))
- Shell
- declared (2 observation(s))
- Dependencies
- pinned
- Secrets in source
- found
Findings (25)
f'(deny process-exec (literal "{_sbpl_quote(path)}"))'f'(deny process-exec (literal "{_sbpl_quote(path)}"))' for path in _RESTRICTED_ESCAPE_EXECUTABLESif name in {".netrc", ".npmrc", ".pypirc", ".git-credentials", "auth.json"}:if "/.aws/credentials" in lower or "/.config/gcloud/application_default_credentials.json" in lower:
if any(marker in lower for marker in ("/.ssh/", "~/.ssh/", ".env", "security find-generic-password", "security find-internet-password", "printenv", " env")):"~/.ssh/", "/.ssh/", ".env", ".netrc", ".npmrc", ".pypirc",
".git-credentials", "credentials.json", "service-account.json",
token: "generated-at-install-time",
TOKEN = "resilience-dashboard-token"
secret = "local-dashboard-secret-0123456789-abcdefghijklmnopqrstuvwxyz"
secret = "TOP_SECRET_SENTINEL_DO_NOT_LEAK"
secret = "SIDEBAR_SECRET_MUST_NOT_LEAK"
secret = "ghp_abcdefghijklmnopqrstuvwxyz123456"
private_key = "-----BEGIN PRIVATE KEY-----\nQUJDREVGR0hJSktMTU5PUFFSU1RVVldYWVo0123456789\n-----END PRIVATE KEY-----"
delete_path
self.assertNotIn("eval(", script)self.assertNotIn("eval(", source)digest = hashlib.sha1(str(native_id).encode("utf-8")).hexdigest()[:16]with patch.object(security.socket, "getaddrinfo", return_value=_dns("169.254.169.254")):For the simplest Cloudflare setup, create the tunnel and hostname in Cloudflare, then paste the tunnel token once into **Settings → Advanced → Cloudflare**. Mac MCP writes it atomically to `~/.mac-mcp
http://127.0.0.1:8000/mcp
| http://127.0.0.1:<port>/mcp
curl --fail http://127.0.0.1:8000/health
http://127.0.0.1:8000/mcp
raw = base64.b64decode(encoded.encode("ascii"), validate=True).decode("utf-8") if encoded else ""Gates applied: no_behavioural_pass.
b6aa4f9a8257full audit observations/trust-audit/mcp-server/bulutarkan__mac-mcp.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 | b6aa4f9a8257 | BLOCK | F | 49 | first audit |
Questions
What is the mac-mcp MCP server?
Talk to your AI agent and let it work on your Mac. Open source MCP control for browser, files, apps, shell, and agents.
What tools does mac-mcp expose?
50 in total: 36 read-only, 13 that write, and 1 that can delete or overwrite (delete_path). Every one is listed on this page with its risk.
Is mac-mcp safe to connect to an agent?
No — not without reading the findings first. The audit graded it F (49/100) and found 7 critical or high issues in the source. Each one is listed on this page with the file and line it is on. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does mac-mcp need?
It reads CLOUDFLARE_TUNNEL_TOKEN_FILE, GROQ_API_KEY, MAC_MCP_CHROME_BRIDGE_TOKEN, MAC_MCP_CHROME_BRIDGE_TOKEN_FILE, MAC_MCP_DASHBOARD_TOKEN_FILE, MAC_MCP_FAKE_PARENT_SECRET, MAC_MCP_OPENROUTER_KEYCHAIN_ACCOUNT, MAC_MCP_OPENROUTER_KEYCHAIN_SERVICE, MAC_MCP_SCOPED_AUTH_DB, MAC_MCP_VOICE_GROQ_API_KEY, MAC_MCP_VOICE_GROQ_DEFAULTS_KEY and MAC_MCP_VOICE_GROQ_KEYCHAIN_ACCOUNT from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does mac-mcp run?
It speaks streamable-http, so it runs as a service you connect to over the network. It is published on PyPI as mac-mcp.
How current is this page?
The grade is for one exact copy of the source (b6aa4f9a8257), read on 2026-10-07. The repository is watched and re-audited when it changes.