Intercept any app, then call it from Python like a library.
from hinge_client import Hinge
acc = Hinge() # reuses your captured session
recs = acc.get_recommendations()
acc.like(subject_id, comment="hi lol")You don't write hinge_client.py. mimic captures your own app traffic and an AI
generates the client from it.
Most apps authenticate every request with the same bundle of values: a bearer token, some device ids, a session id, cookies. They're stable across calls. Capture them once from a real request you made, and you can replay them on new requests to the same API.
capture traffic -> extract auth -> generate client
(mitmproxy) (mimic.Session) (claude reads the
captured endpoints)
The generated client is plain Python on top of mimic.App, and you edit it like
any other file. It gives you named methods, body templates, and the multi-step
call chaining mobile APIs tend to need (fetch a token in one call, spend it in
the next).
sh install.shInstalls uv if you don't have it, then mimic in an
isolated tool env. mitmproxy isn't a separate install; mimic launches it via
uvx on first record. (Manual: uv tool install mimic-client.)
mimic doctor # confirm proxy + claude are readymimic record # starts the proxy, prints the iPhone stepsrecord fills in your Mac's LAN IP and walks you through it:
- iPhone -> Wi-Fi -> Configure Proxy -> Manual ->
<your-mac-ip>:8080 - Safari ->
http://mitm.it-> install the Apple profile - Settings -> General -> About -> Certificate Trust Settings -> turn on full trust for mitmproxy. This step is easy to miss and nothing works without it.
- open the app, use it normally
Then:
mimic hosts # list captured hosts; pick your API host
mimic learn prod-api.hingeaws.net # see the endpoints mimic saw
mimic gen prod-api.hingeaws.net # generate hinge_client.pyThen from hinge_client import Hinge; Hinge().get_recommendations().
Three ways to build a session by hand, if you don't want codegen:
from mimic import Session
Session.from_mitm("prod-api.hingeaws.net") # pull auth from mitmweb
Session.from_curl(open("copied.txt").read()) # paste "Copy as cURL" from devtools
Session(base_url="https://x.com", headers={...}) # explicit.get(path), .post(path, json=...), and the other common HTTP verb helpers
return parsed JSON and raise requests.HTTPError for failed responses. If your
token rotates, a 401 on an idempotent request triggers one re-pull from
mitmweb and a retry. Non-idempotent requests are not retried unless you explicitly
pass refresh=True.
- mitmproxy for iOS apps (the default). mimic reads its JSON flow API and
runs it via
uvx, so there's nothing extra to install. - cURL / paste for anything with a web version.
Copy as cURLin devtools, thenSession.from_curl(text). No proxy, no cert. - HAR file for web apps and anything you can capture in a browser. In Chrome
or Firefox devtools, open the Network tab, right-click a request, and choose
"Save all as HAR". Then
mimic hosts --har traffic.harandmimic gen api.example.com --har traffic.har, or build a session directly withSession.from_har("traffic.har", "api.example.com"). No proxy, no cert.
Learn a connected device's observed HTTP request/response shapes, export OpenAPI for a coding harness, and run a local stand-in after disconnecting the hardware:
mkdir -p .mimic-device
mimic device learn 192.168.1.40 --har device.har \
-o .mimic-device/device.json --openapi .mimic-device/openapi.json
mimic device serve .mimic-device/device.jsonFor direct recording, use mimic.RecordingSession; existing Session/App
clients can use base_url="http://127.0.0.1:8090". Exact captured calls replay
offline; optional explicit rules model state updates. Unmatched calls report
unknown behaviour instead of generating plausible-looking results.
This supports HTTP device APIs, including APIs over USB networking. USB HID, serial, Bluetooth and OS device enumeration need separate adapters. See device emulation guide for examples, harness usage, contract semantics and limitations.
If software refuses to load without a device/package, first identify the interface
it checks. mimic startup capture passively observes a startup window on that
machine; analyze and compare identify network candidates with the device
present and absent. startup learn builds selected plaintext TCP/UDP replay
profiles, and startup serve runs the stand-in before the software starts.
mimic startup interfaces
mimic startup capture --interface INTERFACE --seconds 30 -o .mimic-device/startup.json
mimic startup analyze .mimic-device/startup.json --machine MACHINE_IP \
-o .mimic-device/startup-report.jsonLearning expected replies needs a successful baseline or a documented contract. Network traffic cannot establish package installation or OS USB device presence; those need separate shims/tracing. See the startup emulation guide for the C/M/A workflow, TCP/UDP learning/replay, endpoint mapping and a runnable fictional startup gate.
Two auth schemes get in the way, for different reasons:
- Certificate pinning (banking, Instagram). The app rejects the mitmproxy
cert, so the proxy sees no traffic and nothing shows up in
mimic hosts. This blocks capture, not replay — get past the pin and the rest works normally.mimic unpin <ipa|bundle-id>sets up a Frida-based bypass; see docs/pinning.md. - DPoP / sender-constrained tokens. Each request carries a fresh proof signed by a private key that never leaves the device, so captured requests don't replay. This defeats the core model, not just capture; there's no clean workaround. See docs/dpop.md.
If mimic hosts shows the app's API host, you're good.
Use it on your own accounts and data. It replays your session; it is not a tool for accessing anyone else's. Respect each app's terms of service.
MIT, see LICENSE. Provided as-is, no warranty. Use on your own accounts and data; you are responsible for complying with each app's terms.
