proxy-mcp is an HTTP/HTTPS MITM proxy MCP server for intercepting, modifying, and analyzing network traffic, with advanced TLS fingerprinting, browser automation, and multi-platform device routing.
Core Proxy
Start/stop the proxy with auto-generated CA certificates; retrieve PEM/SPKI fingerprint for device trust installation
Capture traffic in memory (1000 entries) or persistent on-disk sessions
List, search, filter, and inspect captured HTTP exchanges (URL, hostname, method, status, full-text body search)
Interception Rules
Add/update/remove/enable/disable rules with matchers (URL, hostname, method, headers, body) and handlers (passthrough, mock, forward, drop)
Transform traffic in-flight: inject/delete headers, rewrite URLs, replace bodies, change status codes
Test rules against simulated or captured exchanges with per-field diagnostics
TLS Fingerprinting & Spoofing
Capture JA3/JA4 client and JA3S server fingerprints per exchange
Spoof outgoing TLS ClientHello, HTTP/2 frames, and header order using browser presets (Chrome 131/136, Firefox 133+) via native Rust impit
Normalize UA Client Hints to match spoofed User-Agent
Upstream Proxy Chaining
Global or per-host upstream proxies (SOCKS4/5, HTTP, HTTPS, PAC) for geo-routing, authentication, or IP reputation
Bypass lists (
no_proxy) for both global and per-host configs
Browser Interception (cloakbrowser)
Launch stealth-patched Chromium with proxy flags and CA trust auto-configured
Navigate pages, take ARIA/role YAML snapshots and screenshots, read console messages
List/retrieve cookies, localStorage, sessionStorage, and proxy-captured network headers
TLS ClientHello passthrough: browser's native ClientHello forwarded upstream for authentic Chrome TLS fingerprints
Human-like Browser Automation
Mouse movement via Bezier curves with Fitts's law velocity scaling
Click by CSS selector, ARIA role/name, visible text, or form label (with auto-wait)
Typing with WPM-based timing, bigram frequency modeling, optional typos with backspace correction
Scrolling with easeInOutQuad acceleration; idle simulation with micro-jitter and micro-scrolls
Terminal / Process Interception
Spawn processes with proxy env vars pre-configured (HTTP_PROXY, HTTPS_PROXY, SSL_CERT_FILE, NODE_EXTRA_CA_CERTS, CURL_CA_BUNDLE, and 15+ more)
Android Interception
Full device interception via ADB: CA cert injection into system store, ADB reverse tunnel, optional Wi-Fi proxy
Frida-based SSL unpinning: attach to running apps with unpinning + proxy redirect scripts, bypassing certificate pinning
Docker Interception
Inject proxy env vars and CA cert into running containers (live exec or stop+restart mode)
Persistent Sessions & HAR
Start/stop/query persistent sessions with preview or full-body capture
Import HAR files for analysis and replay; export sessions to HAR
Replay recorded requests in dry-run or execute mode with optional target URL override
Recover/rebuild session indexes after crash or corruption
Provides comprehensive traffic interception for Android devices, including CA certificate injection via ADB, reverse tunneling, and SSL unpinning using Frida scripts.
Enables automated launching of the Brave browser with integrated proxy configuration for capturing and modifying web traffic.
Automates proxy configuration for curl commands by setting environment variables to route traffic through the MITM proxy for inspection.
Supports intercepting Deno network traffic by automatically configuring proxy environment variables for spawned processes.
Injects proxy settings and CA certificates into running Docker containers to capture and inspect containerized network traffic.
Enables monitoring and manipulation of Git network requests by automatically configuring proxy environment variables.
Allows for the interception of Node.js network traffic by configuring proxy environment variables and SSL certificate trust for spawned processes.
Routes npm network activity through the proxy by setting appropriate environment variables to capture and inspect package management traffic.
Enables interception of Python application traffic, including support for the requests library, through automated environment variable injection.
Routes yarn network activity through the proxy by setting appropriate environment variables to capture and inspect package management traffic.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@proxy-mcpintercept traffic to api.example.com and show the headers"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
proxy-mcp
proxy-mcp is an MCP server that runs an explicit HTTP/HTTPS MITM proxy (L7). It captures requests/responses, lets you modify traffic in-flight (headers/bodies/mock/forward/drop), supports upstream proxy chaining, and records TLS fingerprints for connections to the proxy (JA3/JA4) plus optional upstream server JA3S. Ships "interceptors" to route stealth browsers (cloakbrowser and Camoufox), CLI tools, Docker containers, and Android devices/apps through the proxy, plus Playwright-driven browser automation with locator-based click, typing, scroll, and ARIA snapshots.
74 tools + 8 resources + 3 resource templates. Built on mockttp, cloakbrowser, and Camoufox.
Table of Contents
Related MCP server: Android Proxy MCP
Setup
Quick install (Claude Code)
claude mcp add proxy-mcp -- npx -y proxy-mcp@latestThis installs proxy-mcp as an MCP server using stdio transport. It auto-updates on every Claude Code restart.
Scopes:
# Per-user (available in all projects)
claude mcp add --scope user proxy-mcp -- npx -y proxy-mcp@latest
# Per-project (shared via .mcp.json, commit to repo)
claude mcp add --scope project proxy-mcp -- npx -y proxy-mcp@latestPrerequisites
Node.js 20+
From source (development)
git clone https://github.com/yfe404/proxy-mcp.git
cd proxy-mcp
npm install
npm run build# stdio transport (default) — used by MCP clients like Claude Code
node dist/index.js
# Streamable HTTP transport — exposes /mcp endpoint for scripting
node dist/index.js --transport http --port 3001--transport and --port also accept env vars TRANSPORT and PORT.
Manual MCP configuration
The configured server alias controls Claude's generated tool prefix. The examples below use proxy-mcp, so Claude Code exposes tools as mcp__proxy-mcp__<tool_name>. If you rename the server key to proxy, use mcp__proxy__<tool_name> instead.
Claude Code CLI:
# stdio (default)
claude mcp add proxy-mcp -- npx -y proxy-mcp@latest
# From local clone
claude mcp add proxy-mcp -- node /path/to/proxy-mcp/dist/index.js
# HTTP transport for scripting
claude mcp add --transport http proxy-mcp http://127.0.0.1:3001/mcp.mcp.json (project-level, commit to repo):
{
"mcpServers": {
"proxy-mcp": {
"command": "npx",
"args": ["-y", "proxy-mcp@latest"]
}
}
}Streamable HTTP transport:
{
"mcpServers": {
"proxy-mcp": {
"type": "streamable-http",
"url": "http://127.0.0.1:3001/mcp"
}
}
}HTTP Proxy Configuration
1) Start proxy and get endpoint
proxy_startUse the returned port and endpoint http://127.0.0.1:<port>.
2) Browser setup (recommended: interceptor)
Use the browser interceptor so proxy flags and cert trust are configured automatically. Launches cloakbrowser — a stealth-patched Chromium with source-level C++ fingerprint patches and humanize mode on by default:
interceptor_browser_launch --url "https://example.com"Drive the page with Playwright-backed tools (no CDP, no sidecar — target_id is all you need):
interceptor_browser_navigate --target_id "browser_<id>" --url "https://apify.com"
interceptor_browser_snapshot --target_id "browser_<id>"
interceptor_browser_screenshot --target_id "browser_<id>" --file_path "/tmp/shot.png"3) Browser setup (manual fallback)
If launching a browser manually, pass the proxy flag yourself:
google-chrome --proxy-server="http://127.0.0.1:<port>"4) CLI/process setup
Route any process through proxy-mcp by setting proxy env vars:
export HTTP_PROXY="http://127.0.0.1:<port>"
export HTTPS_PROXY="http://127.0.0.1:<port>"
export NO_PROXY="localhost,127.0.0.1"If the client verifies TLS, trust the proxy-mcp CA certificate (see proxy_get_ca_cert) or use the Terminal interceptor (interceptor_spawn) which sets proxy env vars plus common CA env vars (curl, Node, Python requests, Git, npm/yarn, etc.):
interceptor_spawn --command curl --args '["-s","https://example.com"]'Explicit curl examples:
curl --proxy http://127.0.0.1:<port> http://example.com
curl --proxy http://127.0.0.1:<port> https://example.com5) Upstream proxy chaining
Set optional proxy chaining from proxy-mcp to another upstream proxy (for geolocation, auth, or IP reputation):
Client/app → proxy-mcp (local explicit proxy) → upstream proxy (optional chaining layer)proxy_set_upstream --proxy_url "socks5://user:pass@upstream.example:1080"Supported upstream URL schemes: socks4://, socks5://, http://, https://, pac+http://.
Typical geo-routing examples:
# Route ALL outgoing traffic from proxy-mcp via a geo proxy
proxy_set_upstream --proxy_url "socks5://user:pass@fr-exit.example.net:1080"
# Bypass upstream for local/internal hosts
proxy_set_upstream --proxy_url "http://user:pass@proxy.example.net:8080" --no_proxy '["localhost","127.0.0.1",".corp.local"]'
# Route only one hostname via a dedicated upstream (overrides global)
proxy_set_host_upstream --hostname "api.example.com" --proxy_url "https://user:pass@us-exit.example.net:443"
# Remove overrides when done
proxy_remove_host_upstream --hostname "api.example.com"
proxy_clear_upstreamFor HTTPS MITM, the proxy CA must be trusted in the target environment (proxy_get_ca_cert).
6) Validate and troubleshoot quickly
proxy_list_traffic --limit 20
proxy_search_traffic --query "example.com"Common issues:
Traffic from the wrong browser instance (fix: always pass
target_idfrominterceptor_browser_launch)HTTPS cert trust missing on target
NO_PROXYbypassing expected hostsFirst launch is slow: cloakbrowser downloads a ~200 MB stealth Chromium binary on first use (cached afterwards)
7) HAR import + replay
Import HAR into a persisted session, then analyze with existing session query/findings tools:
proxy_import_har --har_file "/path/to/capture.har" --session_name "imported-run"
proxy_list_sessions
proxy_query_session --session_id SESSION_ID --hostname_contains "api.example.com"
proxy_get_session_handshakes --session_id SESSION_IDReplay defaults to dry-run (preview only). Execute requires explicit mode:
# Preview what would be replayed
proxy_replay_session --session_id SESSION_ID --mode dry_run --limit 20
# Execute replay against original hosts
proxy_replay_session --session_id SESSION_ID --mode execute --limit 20
# Optional: override target host/base URL while preserving path+query
proxy_replay_session --session_id SESSION_ID --mode execute --target_base_url "http://127.0.0.1:8081"Note: imported HAR entries (and entries created by proxy_replay_session) do not carry JA3/JA4/JA3S handshake metadata. Use live proxy-captured traffic to analyze handshake fingerprints.
Mobile Capture (Transparent Proxy)
For mobile apps with custom HTTP stacks that ignore the system proxy (most modern Android apps — Shopee, SHEIN, TikTok, banking, etc.) the explicit proxy won't see their traffic. proxy-mcp ships a transparent listener that sits behind an iptables REDIRECT and MITMs using the TLS SNI — no CONNECT tunnel required.
Pairs with proxy-ap-card — a XIAO ESP32-S3 that broadcasts a WiFi AP (proxy-ap SSID by default) and presents to the laptop as a USB-NCM ethernet adapter. That repo handles the AP + NAPT side; proxy-mcp handles the laptop side.
Prerequisites
Laptop: Linux with
iptables,sysctl,nmcli(NetworkManager),ip(iproute2),adb.sudofor network configuration (one command per session).Hardware router: a proxy-ap-card (XIAO ESP32-S3) with firmware flashed. Or any USB-ethernet / USB-WiFi combo where the laptop NATs traffic for the phone's subnet — pass
--ap_iface/--ap_subnetto override defaults.Target device: Android with root (Magisk / KernelSU). Root is required to inject the CA into the system cert store. Android 14-16 additionally need the zygote mount-namespace injection which this tool does automatically.
ADB access: the target device must appear in
adb devices(USB at minimum, wireless after pairing).
First-time walkthrough
1. Flash + plug in the proxy-ap-card
Follow the proxy-ap-card README to build + flash the XIAO. On replug the laptop should show a cdc_ncm interface (verify: ls /sys/class/net/*/device/uevent | xargs grep DRIVER | grep cdc_ncm).
2. Connect the target device to the laptop via USB
Only needed the first time per device, to push the CA. Verify with adb devices. Copy the serial.
3. Run the setup tool
proxy_mobile_setup --android_serial <serial>Optional arguments:
Param | Default | When to override |
| auto-detect ( | Using a different USB-ethernet bridge |
|
| Matches default proxy-ap-card firmware |
|
| Matches default proxy-ap-card firmware |
| auto-detect from default route | Multi-homed host, or wanting traffic to exit via a specific iface |
|
| Port collision |
|
| Port collision |
|
| Need QUIC/HTTP3 (no MITM available then) |
| — | Route outbound through a residential/ISP proxy — see below |
| — | Omit to skip CA injection (remote-only setups) |
|
| Set |
The response is JSON with three key bits:
{
"ap_iface": "enp195s0f3u1u4",
"cert_injected": true,
"android_target_id": "adb_HQ63C81CB2",
"sudo_command": "sudo bash /tmp/proxy-mcp-mobile-setup-<hex>.sh"
}Keep android_target_id around — you'll need it for teardown.
4. Run the emitted sudo script
sudo bash /tmp/proxy-mcp-mobile-setup-<hex>.shThis is the only sudo required. The script is idempotent, safe to re-run. It plumbs the routing that iptables needs root for:
# Static IP on the AP iface (skipped if already configured).
ip addr add 192.168.99.2/24 dev <ap_iface>
# Forwarding + dedicated nat chain.
sysctl -w net.ipv4.ip_forward=1
iptables -t nat -N PROXY_MCP_PREROUTING
iptables -t nat -A PROXY_MCP_PREROUTING -p tcp --dport 80 -j REDIRECT --to-ports 8080
iptables -t nat -A PROXY_MCP_PREROUTING -p tcp --dport 443 -j REDIRECT --to-ports 8443
iptables -t nat -A PREROUTING -i <ap_iface> -j PROXY_MCP_PREROUTING
# Block QUIC so apps fall back to TCP/TLS.
iptables -A FORWARD -i <ap_iface> -p udp --dport 443 -j DROP
# Masquerade out through the real egress.
iptables -t nat -A POSTROUTING -s 192.168.4.0/24 -o <egress_iface> -j MASQUERADEWhy a script and not direct execution? MCP runs as your user; iptables needs root; sudo from an MCP tool would require NOPASSWD or polkit policy (fragile, distro-specific). Emitting a script is auditable, reproducible, and portable.
5. Connect the phone to the proxy-ap WiFi
Credentials are set in the proxy-ap-card firmware (default SSID proxy-ap, default password shown in that repo). After connecting, the phone can be unplugged from USB — future sessions don't need it.
6. Start capturing
Every HTTP/HTTPS request from the phone now lands in proxy-mcp's ring buffer:
proxy_list_traffic --source_filter transparent # iptables-redirected HTTPS
proxy_list_traffic --source_filter explicit # absolute-URL HTTP that arrived on :80
proxy_get_exchange --exchange_id <id> # full headers + body preview
proxy_search_traffic --query "api.example.com" # full-text searchEach entry carries source: "explicit" | "transparent" + TLS fingerprints (ja3, ja4).
Subsequent sessions (phone already paired)
Skip step 2. Run:
proxy_mobile_setup # skip android_serial — cert already installed
sudo bash /tmp/proxy-mcp-mobile-setup-<hex>.shThen phone joins the AP and capture resumes.
Upstream chaining
Set upstream_proxy_url to route outbound traffic through a residential/ISP proxy — the target servers see that proxy's IP, not your laptop's:
proxy_mobile_setup \
--upstream_proxy_url "http://user:pass@proxy.example.com:8000" \
--android_serial <serial>Applies to BOTH listeners. Use proxy_set_upstream after the fact to change it without restarting.
Verifying each step
Check | Command | Expected |
Listeners up |
|
|
Iface detected |
|
|
Cert injected |
| 144 (or 143 + 1) |
iptables wired |
| 2 REDIRECT rules, non-zero |
Forwarding on |
|
|
Phone sees AP | phone Settings → WiFi shows | |
Traffic flowing |
|
|
Teardown
proxy_mobile_teardown --android_target_id <from-setup>
sudo bash /tmp/proxy-mcp-mobile-teardown-<hex>.shThe MCP call stops both listeners and deactivates the Android target. The sudo script removes the iptables rules, disables ip_forward, and hands the AP iface back to NetworkManager. Phone stays connected to the AP until you forget it in phone WiFi settings.
Troubleshooting
Symptom | Likely cause | Fix |
| proxy-ap-card not plugged in, or |
|
| iface name passed explicitly but doesn't exist | Use auto-detect or verify with |
Phone connects to AP but zero traffic in | sudo script not run; or phone on a different WiFi | Verify |
Phone says "no internet" on | Forward + MASQUERADE rules missing, OR egress iface down | Re-run the sudo script; check |
HTTPS fails with "connection not private" or similar | Cert not trusted by the app (Chrome bundles its own CAs and ignores system trust, app has cert pinning) | Use another app to verify chain works; Chrome is the exception not the rule (see limitations below) |
Some apps capture, others don't | Cert pinning in the apps that fail | See limitations below; Frida/LSPosed unpinning module needed |
Ports 20346/20443/other custom ports not captured | Only | Add extra REDIRECT rules in the sudo script, or pair redsocks to |
| Device not rooted, or SELinux |
|
Wireless ADB port changes every session | Android's Wireless Debugging randomises the port | Re-pair; or keep phone plugged in for control plane |
Without the proxy-ap-card
Any USB-NCM or USB-ethernet bridge that the phone can route through works:
proxy_mobile_setup --ap_iface eth1 --ap_address 10.0.0.1/24 --ap_subnet 10.0.0.0/24Or use a laptop-hosted WiFi AP via hostapd on a USB WiFi adapter (MT76x2U, RTL8812AU, etc.) — pass the wlanN interface as ap_iface.
Limitations
Cert pinning — apps that pin specific public keys (Instagram, WhatsApp, banking, Shopee homepage feed, etc.) will refuse our mockttp cert even with CA trust installed. You see partial capture (tracking / static / unpinned endpoints succeed, pinned API calls fail). Fix: Frida or LSPosed unpinning module for that specific app. See
interceptor_frida_attachin the tool reference, or morrownr's USB-WiFi guides for common patterns.Chrome on Android — Chrome ships its own Mozilla CA bundle and enforces Certificate Transparency. Our CA is self-signed and not in any CT log, so Chrome rejects it regardless of system trust. Use any other app (or any OkHttp/Conscrypt-based browser) to verify the pipeline.
QUIC / HTTP3 — the transparent listener is TCP/TLS only. By default we drop UDP/443 so apps fall back to TCP. Set
block_quic: falseif you want QUIC to pass through uncaptured (QUIC content won't appear in traffic logs).Non-standard ports — the default iptables rules only redirect TCP/80 and TCP/443. Shopee's
:20346, custom gaming protocols, etc. will bypass. Add more REDIRECT rules in the sudo script (or chainredsocksto issue CONNECT tunnels through the explicit listener).Native TLS pinning — some apps use BoringSSL/OpenSSL directly via JNI with pins embedded in the
.so. Java-layer Frida hooks won't catch these; native hooks needed.Root required — the system-trust CA overlay requires root. Non-rooted Android only trusts user-installed CAs for apps that explicitly opt in via
network_security_config— which no shipping app does. There's no bypass for that without root.
Boundaries
Only sees traffic configured to route through it (not a network tap or packet sniffer)
Spoofs outgoing JA3 + HTTP/2 fingerprint + header order (via impit — native Rust TLS impersonation), not JA4 (JA4 is capture-only)
Can add, overwrite, or delete HTTP headers; outgoing header order can be controlled via fingerprint spoofing
Returns its own CA certificate — does not expose upstream server certificate chains
TLS ClientHello Passthrough (browser via interceptor)
When cloakbrowser is launched via interceptor_browser_launch, proxy-mcp forwards the browser's original TLS ClientHello to the upstream server for document loads and same-origin sub-resource requests. The target server sees an authentic Chrome TLS fingerprint — not the proxy's.
This is a key difference from typical MITM proxies (mitmproxy, Charles, Fiddler) which re-terminate TLS with their own fingerprint, making MITM trivially detectable by anti-bot systems via JA3/JA4 analysis.
How to verify passthrough is working:
proxy_list_tls_fingerprints --hostname_filter "example.com"JA3 varies across requests to the same host — this is expected; Chrome randomizes cipher suite order per-connection (feature since Chrome 110+)
JA4 stays stable — same cipher/extension set, just different ordering
JA3 variation + JA4 stability = authentic Chrome TLS passthrough confirmed
When passthrough applies vs. when spoofing is needed:
Traffic source | TLS behavior | Action needed |
cloakbrowser via | Browser's native ClientHello forwarded (passthrough) | None — fingerprint is authentic |
cloakbrowser via | Re-issued via impit with spoofed TLS |
|
Non-browser clients (curl, Python, | Proxy's own TLS |
|
HAR replay ( | Proxy's own TLS |
|
Built on stealth browsers + Playwright
Browser automation uses cloakbrowser for stealth-patched Chromium and Camoufox for anti-detect Firefox, both driven through Playwright. There is no CDP sidecar or hand-rolled stealth script in proxy-mcp. Downstream tools take a browser_* target from interceptor_browser_launch or a camoufox_* target from interceptor_camoufox_launch.
Capability | proxy-mcp |
See/modify DOM, run JS in page |
|
Read cookies, localStorage, sessionStorage | Yes — |
Capture HTTP request/response bodies | Via the MITM proxy (4 KB preview cap by default; |
Modify requests in-flight (headers, body, mock, drop) | Yes (declarative rules, hot-reload) |
Upstream proxy chaining (geo, auth) | Global + per-host upstreams across all clients (SOCKS4/5, HTTP, HTTPS, PAC) |
TLS fingerprint capture (JA3/JA4/JA3S) | Yes |
JA3 + HTTP/2 fingerprint spoofing | Proxy-side (impit re-issues matching requests with spoofed TLS 1.3, HTTP/2 frames, and header order) |
Intercept non-browser traffic (curl, Python, Android apps) | Yes (interceptors) |
Human-like mouse/keyboard/scroll input |
|
Locator-based interaction |
|
Standard flow:
Call
proxy_startOptionally enable outbound fingerprint spoofing for cross-origin sub-resources:
proxy_set_fingerprint_spoof --preset chrome_136Call
interceptor_browser_launch --url "https://example.com"orinterceptor_camoufox_launchDrive the page:
interceptor_browser_navigate,interceptor_browser_snapshot,humanizer_click --selector "...",humanizer_type --text "..."Inspect traffic:
proxy_search_traffic --query "<hostname>"
Tools Reference
Lifecycle (4)
Tool | Description |
| Start MITM proxy, auto-generate CA cert |
| Stop proxy (traffic/cert retained) |
| Running state, port, rule/traffic counts |
| CA certificate PEM + SPKI fingerprint |
Transparent / Mobile Capture (6)
Tool | Description |
| Start second MITM listener (SNI-based, no CONNECT) on a parallel port; shares CA + rules + ring buffer with the explicit listener |
| Stop the transparent listener |
| Running state + port + dedicated traffic count |
| One-command mobile capture: start both listeners, inject CA into Android system store via adb (tmpfs overlay + zygote ns for Android 14+), emit a sudo-runnable iptables/sysctl/nmcli script |
| Reverse setup: deactivate Android target, stop transparent listener, emit teardown script |
| Probe |
Upstream Proxy (4)
Tool | Description |
| Set global upstream proxy |
| Remove global upstream |
| Per-host upstream override |
| Remove per-host override |
Interception Rules (7)
Tool | Description |
| Add rule with matcher + handler |
| Modify existing rule |
| Delete rule |
| List all rules by priority |
| Test which rules would match a simulated request or captured exchange, with detailed diagnostics |
| Enable a disabled rule |
| Disable without removing |
Quick debugging examples:
# Simulate a request and see which rule would win
proxy_test_rule_match --mode simulate --request '{"method":"GET","url":"https://example.com/api/v1/items","headers":{"accept":"application/json"}}'
# Evaluate a real captured exchange by ID
proxy_test_rule_match --mode exchange --exchange_id "ex_abc123"Traffic Capture (4)
Tool | Description |
| Paginated traffic list with filters |
| Full exchange details by ID |
| Full-text search across traffic |
| Clear capture buffer |
Modification Shortcuts (3)
Tool | Description |
| Add/overwrite/delete headers on matching traffic (set value to |
| Rewrite request URLs |
| Return mock response for matched requests |
TLS Fingerprinting (9)
Tool | Description |
| Get JA3/JA4 client fingerprints + JA3S for a single exchange |
| List unique JA3/JA4 fingerprints across all traffic with counts |
| Legacy: enable JA3 spoofing (deprecated, use |
| Disable fingerprint spoofing |
| Return current TLS config (server capture, JA3 spoof state) |
| Toggle server-side JA3S capture (monkey-patches |
| Enable full TLS + HTTP/2 fingerprint spoofing via impit. Supports browser presets. |
| List available browser fingerprint presets (e.g. |
| Check fingerprint spoofing backend readiness |
Fingerprint spoofing works by re-issuing the request from the proxy via impit (native Rust TLS/HTTP2 impersonation via rustls). TLS 1.3 and HTTP/2 fingerprints (SETTINGS, WINDOW_UPDATE, PRIORITY frames) match real browsers by construction. The origin server sees the proxy's spoofed TLS, HTTP/2, and header order — not the original client's. When a user_agent is set (including via presets), proxy-mcp also normalizes Chromium UA Client Hints headers (sec-ch-ua*) to match the spoofed User-Agent (forwarding contradictory hints is a common bot signal). Browser exception: when cloakbrowser is launched via interceptor_browser_launch, document loads and same-origin requests use the browser's native TLS (no impit), preserving fingerprint consistency for bot detection challenges. Only cross-origin sub-resource requests are re-issued with spoofed TLS. Non-browser clients (curl, spawn, HAR replay) get full TLS + UA spoofing on all requests. Use proxy_set_fingerprint_spoof with a browser preset for one-command setup. proxy_set_ja3_spoof is kept for backward compatibility but custom JA3 strings are ignored (the preset's impit browser target is used instead). JA4 fingerprints are captured (read-only) but spoofing is not supported.
Interceptors (21)
Interceptors configure targets (browsers, processes, devices, containers) to route their traffic through the proxy automatically.
Discovery (3)
Tool | Description |
| List all interceptors with availability and active target counts |
| Detailed status of a specific interceptor |
| Emergency cleanup: kill all active interceptors across all types |
Browser (3)
Tool | Description |
| Launch cloakbrowser (stealth Chromium) with proxy flags, SPKI cert trust, built-in humanize mode |
| Navigate the bound page via Playwright |
| Close a browser instance by target ID |
Stealth is source-level: cloakbrowser ships 48+ C++ patches so ja3n/ja4/akamai match real Chrome, navigator.webdriver is false, audio/canvas/WebGL fingerprints match real hardware. No JS stealth injection needed. First launch downloads a ~200 MB Chromium binary (cached afterwards).
Camoufox (4) — anti-detect Firefox
Tool | Description |
| Spawn camoufox as a Playwright WebSocket server, proxy + NSS CA pre-wired. Returns |
| Get the wsUrl + ready-to-paste TS / Python |
| List active camoufox instances and their fingerprint details |
| Stop the launcher, remove the temp launcher dir + NSS profile |
Camoufox is a patched Firefox with source-level fingerprint controls (OS, WebGL vendor/renderer, fonts, locale, geoip-derived timezone, WebRTC blocking, humanize cursor). Camoufox runs as an external Python process and exposes a Playwright WS endpoint, but proxy-mcp also binds the returned camoufox_* target to the same interceptor_browser_* and humanizer_* tools used by cloakbrowser. Use the exposed wsUrl only when you need custom Playwright code outside MCP.
By default, proxy-mcp sets Camoufox fingerprint generation to the host OS (linux, macos, or windows) instead of Camoufox's upstream random OS list. Override with os when you intentionally need a different family, or pass an array such as ["windows", "macos"] to let Camoufox choose from that subset. Launch/list/info responses include a safe fingerprint summary with the resolved OS, User-Agent, platform, OSCPU, screen/window dimensions, WebGL vendor/renderer, and font/voice counts; raw Camoufox config and process environment are not exposed.
Host requirements:
pip install "cloverlabs-camoufox[geoip]" # active fork; daijro/camoufox stale on Firefox 135 → DataDome distrusts
python3 -m camoufox fetch official/150.0.2-alpha.26 # Firefox 150; default `fetch` still picks v135 due to repos.yml constraint
# For TLS MITM trust (NSS profile is created per-launch and the proxy CA is imported):
sudo apt install libnss3-tools # Debian/Ubuntu
sudo dnf install nss-tools # Fedora/RHEL
# macOS: brew install nss (or use /Applications/Firefox.app/Contents/MacOS/certutil)If certutil is missing, the launch still succeeds but the proxy CA is not trusted — HTTPS pages will show certificate errors. Proxy traffic is still captured.
Usage:
proxy_start // start the MITM proxy
interceptor_camoufox_launch { headless: true } // returns { targetId, wsUrl, fingerprint, ... }
interceptor_browser_navigate --target_id "camoufox_<id>" --url "https://example.com"
interceptor_browser_snapshot --target_id "camoufox_<id>" --mode ai
interceptor_browser_list_console --target_id "camoufox_<id>"
interceptor_browser_list_cookies --target_id "camoufox_<id>"
// Or in your own Node code:
// import { firefox } from 'playwright-core';
// const browser = await firefox.connect(wsUrl);
// const page = await (await browser.newContext()).newPage();
// await page.goto('https://example.com');
interceptor_camoufox_close { target_id } // when doneplaywright-core is already a proxy-mcp dependency — Camoufox uses its firefox namespace via WebSocket; no extra Node packages needed. Traffic capture, TLS fingerprinting, rules, mocks, sessions, upstream chaining, and JA3/JA4 spoofing all apply to camoufox automatically because the proxy sits in front of it.
Terminal / Process (2)
Tool | Description |
| Spawn a command with proxy env vars pre-configured (HTTP_PROXY, SSL certs, etc.) |
| Kill a spawned process and retrieve stdout/stderr |
Sets 18+ env vars covering curl, Node.js, Python requests, Deno, Git, npm/yarn.
Android ADB (4)
Tool | Description |
| List connected Android devices via ADB |
| Full interception: inject CA cert, ADB reverse tunnel, optional Wi-Fi proxy |
| Remove ADB tunnel and clear Wi-Fi proxy |
| Quick setup: push CA cert + ADB reverse tunnel (no Wi-Fi proxy) |
Caveats: CA cert injection requires root access. Supports Android 14+ (/apex/com.android.conscrypt/cacerts/). Wi-Fi proxy is opt-in (default off).
Android Frida (3)
Tool | Description |
| List running apps on device via Frida |
| Attach to app and inject SSL unpinning + proxy redirect scripts |
| Detach Frida session from app |
Caveats: Requires frida-server running on device. Uses frida-js (pure JS, no native binaries on host). SSL unpinning covers OkHttp, BoringSSL, TrustManager, system TLS — but may not work against QUIC or custom TLS stacks.
Docker (2)
Tool | Description |
| Inject proxy env vars and CA cert into running container |
| Remove proxy config from container |
Two modes: exec (live injection, existing processes need restart) and restart (stop + restart container). Uses host.docker.internal for proxy URL.
Browser DevTools-equivalents (12)
Playwright-driven tools for the browser target. Each takes a target_id directly — no session binding, no sidecar. Works on both cloakbrowser (browser_* IDs) and camoufox (camoufox_* IDs) targets via the shared getPageForTarget() resolver.
Tool | Description |
| ARIA/role YAML snapshot of the page (or selector subtree) — optimized for LLM page reasoning |
| Screenshot. Writes to |
| Buffered console messages since launch, with type/text filters and pagination |
| Cookie listing with filters, pagination, truncated value previews |
| Get one cookie by |
| localStorage/sessionStorage key listing with value previews |
| Get one storage value by |
| Header field listing from proxy-captured traffic since the browser was launched |
| Get one full header field value by |
| Run a JS file in the page (file body wrapped as |
| Inject a JS file as |
| Append a |
Network data is sourced from the MITM proxy rather than a browser-side protocol — the proxy sees every wire request regardless of what the browser reported.
Stealth tradeoffs for JS injection:
Method | Cloakbrowser | Camoufox (cloverlabs/FF150) |
| Safe (isolated utility world) — rate-limit before reCAPTCHA, each call is CDP traffic | Runs in page main world; reads are invisible, mutations are page-observable |
| Not supported by Playwright API | Requires |
| Best for stealth — pre-document, no DOM artifact | Patches reach the page (good) but are observable via |
| Detectable (DOM node, MutationObserver, CSP) | Detectable (same) |
References: Playwright evaluate, Playwright addInitScript, Camoufox stealth.
Worlds and isolation — what your JS can and can't see
The two backends ship different world models. Picking the wrong tool is the most common stealth footgun, so the boundary matters.
Cloakbrowser (Chromium). Playwright's evaluate runs in an isolated "utility" world that shares globals with the page's main world. An addInitScript patch to navigator.webdriver is visible to (a) your subsequent evaluate probes AND (b) anti-bot code the site loads. This is the model most "stealth playbooks" assume. Detection vectors are CDP-side (Runtime.evaluate chatter) — cloakbrowser's C++ patches mitigate those.
Camoufox (cloverlabs ≥0.6 + Firefox 150 — current build). No separate JS world. page.evaluate and page.addInitScript both run in the page's main world, the same realm as a real <script> tag. Implications:
inject_init_scriptpatches reach the page (e.g.Object.defineProperty(navigator, 'webdriver', ...)does affect what site scripts see). The DOWNSIDE: the patch is observable to anti-bot code on the page —Function.prototype.toString.toString()reveals replaced functions,Object.definePropertyhooks see the call.interceptor_browser_evaluatereads are invisible (nowindowwrites, no prototype changes). Mutating evals (() => { window.x = 1 }) are observable.world: "isolated"andworld: "main"accept the same script args for API compatibility but run in the same realm oncemainis enabled.main_world_eval: truestill gates explicitworld: "main"calls in proxy-mcp; on this build it does not create a separate execution realm.
Verify behavior on your installed build:
npx tsx scripts/camoufox-world-probe.ts --venv=/path/to/camoufox-venvHistorical note. Earlier daijro/camoufox (Firefox 135 line) ran evaluate and addInitScript in a separate Juggler scope that was invisible to the page — patches there did NOT reach site scripts (camoufox#48), but automation JS was equally invisible to anti-bot code. Cloverlabs/FF150 dropped that isolation. If your workflow depends on Juggler-scope invisibility, stay on daijro/FF135.
Practical rules:
Use case | Cloakbrowser | Camoufox (cloverlabs/FF150) |
Read DOM / extract data |
|
|
Modify page state, click via JS |
|
|
Spoof navigator / window fingerprints |
| Configure at launch ( |
Load a 3rd-party JS lib into the page |
| Same — runs in main world; DOM node is detectable |
Stealth note: on the current camoufox build, every JS-level mutation from automation is observable by anti-bot code on the page. Prefer source-level configuration over runtime patching. Use evaluate for reads, not writes, when stealth matters.
Sessions (14)
Persistent, queryable on-disk capture for long runs and post-crash analysis.
Tool | Description |
| Start persistent session capture (preview or full-body mode) |
| Stop and finalize the active persistent session |
| Runtime status for persistence (active session, bytes, disk cap errors) |
| Import a HAR file from disk into a new persisted session |
| List recorded sessions from disk |
| Get manifest/details for one session |
| Indexed query over recorded exchanges |
| Search request/response bodies stored in a persistent session, with context snippets |
| Report JA3/JA4/JA3S handshake metadata availability for session entries |
| Fetch one exchange from a session (with optional full bodies) |
| Dry-run or execute replay of selected session requests |
| Export full session or filtered subset to HAR |
| Delete a stored session |
| Rebuild indexes from records after unclean shutdown |
proxy_get_session_exchange and proxy_export_har automatically decompress response bodies (gzip, deflate, brotli) based on the stored content-encoding header. The returned responseBodyText and responseBodyBase64 contain the decompressed content. Raw compressed bytes are preserved on disk for exact replay fidelity.
Note on proxy_start with persistence_enabled: true: this auto-creates a session. A subsequent proxy_session_start() call returns the existing active session instead of failing — no need to stop and re-start.
Humanizer — Playwright Input (5)
Human-like browser input via Playwright page.mouse / page.keyboard. Works with browser_* targets from interceptor_browser_launch and camoufox_* targets from interceptor_camoufox_launch. Cloakbrowser's own humanize patches apply when enabled at launch; Camoufox cursor humanization follows the Camoufox launch config.
Tool | Description |
| Move the mouse to |
| Click a locator ( |
| Type text into the focused element via |
| Dispatch one Playwright |
| Simulate idle behavior with mouse micro-jitter and occasional micro-scrolls to defeat idle detection |
All tools require target_id from a prior interceptor_browser_launch or interceptor_camoufox_launch. The engine maintains tracked mouse position across calls for coordinate-based move/click/idle behavior.
Behavioral details:
Mouse:
humanizer_movecallspage.mouse.move; locator clicks call Playwright locators and raw-coordinate clicks callpage.mouse.clickTyping:
humanizer_typecallspage.keyboard.type(text, { delay })whendelay_msis provided; no WPM, typo, or bigram model is implemented in proxy-mcpScrolling:
humanizer_scrollsends one wheel event with the requested deltaIdle: Periodic micro-jitter (±3px subtle / ±8px normal) and random micro-scrolls at configurable intensity
Resources
URI | Description |
| Proxy running state and config |
| CA certificate PEM |
| Traffic stats: method/status breakdown, top hostnames, TLS fingerprint stats |
| All interceptor metadata and activation status |
| Persistent session catalog + runtime persistence status |
| Current page URL/title for the most recently launched browser instance |
| Current page state for all active browser instances |
| Active camoufox instances with their wsUrl and fingerprint details |
| Aggregate stats for one recorded session (resource template) |
| Time-bucketed request/error timeline (resource template) |
| Top errors/slow exchanges/host error rates (resource template) |
Usage Example
# Start the proxy
proxy_start
# Optional: start persistent session recording
proxy_session_start --capture_profile full --session_name "reverse-run-1"
# Configure device to use proxy (Wi-Fi settings or interceptors)
# Install CA cert on device (proxy_get_ca_cert)
# Or use interceptors to auto-configure targets:
interceptor_browser_launch # Launch stealth browser with proxy
interceptor_spawn --command curl --args '["https://example.com"]' # Spawn proxied process
interceptor_android_activate --serial DEVICE_SERIAL # Android device
# Set upstream proxy for geolocation
proxy_set_upstream --proxy_url socks5://user:pass@geo-proxy:1080
# Mock an API response
proxy_mock_response --url_pattern "/api/v1/config" --status 200 --body '{"feature": true}'
# Inject auth headers (set value to null to delete a header)
proxy_inject_headers --hostname "api.example.com" --headers '{"Authorization": "Bearer token123"}'
# View captured traffic
proxy_list_traffic --hostname_filter "api.example.com"
proxy_search_traffic --query "error"
# TLS fingerprinting
proxy_list_tls_fingerprints # See unique JA3/JA4 fingerprints
proxy_set_ja3_spoof --ja3 "771,4865-..." # Spoof outgoing JA3 (for non-browser clients)
proxy_set_fingerprint_spoof --preset chrome_136 --host_patterns '["example.com"]' # Full fingerprint spoof
proxy_list_fingerprint_presets # Available browser presets
# Human-like browser interaction (browser_* or camoufox_* target)
humanizer_move --target_id "browser_<id>" --x 500 --y 300
humanizer_click --target_id "browser_<id>" --selector "#login-button"
humanizer_click --target_id "browser_<id>" --role "button" --name "Sign in"
humanizer_type --target_id "browser_<id>" --text "user@example.com" --delay_ms 45
humanizer_scroll --target_id "browser_<id>" --delta_y 300
humanizer_idle --target_id "browser_<id>" --duration_ms 2000 --intensity subtle
# Run / inject JS in the page (cloakbrowser + camoufox)
interceptor_browser_evaluate --target_id "browser_<id>" --script_path /tmp/probe.js
interceptor_browser_evaluate --target_id "camoufox_<id>" --script_path /tmp/probe.js --world main # camoufox + main_world_eval=true
interceptor_browser_inject_init_script --target_id "browser_<id>" --script_path /tmp/hook.js # applies on next navigation
interceptor_browser_add_script_tag --target_id "browser_<id>" --script_path /tmp/lib.js # DOM-visible — avoid for stealth
# Query/export recorded session
proxy_list_sessions
proxy_query_session --session_id SESSION_ID --hostname_contains "api.example.com"
proxy_export_har --session_id SESSION_IDArchitecture
State:
ProxyManagersingleton manages mockttp server, rules, trafficRule rebuild: Rules must be set before mockttp
start(), so rule changes trigger stop/recreate/restart cycleTraffic capture:
on('request')+on('response')events, correlated by request IDRing buffer: 1000 entries max, body previews capped at 4KB
TLS capture: Client JA3/JA4 from mockttp socket metadata; server JA3S via
tls.connectmonkey-patchTLS spoofing: impit (native Rust TLS/HTTP2 impersonation via rustls); in-process, no container needed
Interceptors: Managed by
InterceptorManager, each type registers independentlyBrowser: cloakbrowser (stealth Chromium, ~200 MB binary auto-downloaded on first launch) and Camoufox (anti-detect Firefox, installed separately) driven via Playwright
BrowserContext/PageHumanizer: Singleton engine using Playwright's
page.mouse/page.keyboard, plus local mouse-position tracking for idle jitter
Testing
npm test # All tests (unit + integration)
npm run test:unit # Unit tests only
npm run test:integration # Integration tests
npm run test:e2e # E2E fingerprint tests (requires cloakbrowser + internet)Credits
Core Libraries
Project | Role |
MITM proxy engine, rule system, CA generation | |
Native TLS/HTTP2 fingerprint impersonation (Rust via NAPI-RS) | |
Pure-JS Frida client for Android instrumentation | |
Stealth-patched Chromium with source-level C++ fingerprint patches | |
Anti-detect Firefox backend with source-level fingerprint controls | |
Browser automation API driving cloakbrowser and Camoufox | |
MCP server framework |
Vendored Frida Scripts
All scripts in src/frida-scripts/vendor/ are derived from httptoolkit/frida-interception-and-unpinning (MIT):
config-template.js— proxy/cert config injectionandroid-certificate-unpinning.js— TrustManager + OkHttp + BoringSSL hooksandroid-system-certificate-injection.js— runtime cert injection via KeyStoreandroid-proxy-override.js— ProxySelector monkey-patchnative-tls-hook.js— BoringSSL/OpenSSL native hooksnative-connect-hook.js— libcconnect()redirect
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/yfe404/proxy-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
