Not Backed by Y Combinator Combinator

Crack
any website in minutes.

Scrape Automate Build

Give Traceminer a HAR file (a recording of a website's network traffic), and it reverse engineers the entire API in minutes.

Chat with the HAR to understand the site, find issues, scrape data, run API flows, and export results in any format, no official APIs required.

Trace session
Live analysis
NetworkAuthCompareReplay
MethodEndpointStatus
GET

/api/v2/users/me

Core

200
POST

/oauth/token

Auth

200
GET

/api/v2/products?page=1

Cache

304
PUT

/api/v2/cart/items

Write

200
42% modeled
How It Works

Upload the trace. Let the agents do the hard part.

Uploader

Import a browser session HAR

Ready
upload

Drop your HAR file here

Upload Chrome, Safari, or Firefox exports and preserve headers, cookies, and timing data.

Choose file

Requests

148

Cookies

12

Headers

Auth

Choose mode

Turn one HAR into the workflow you need

Pick the mode that matches what you want to do next.

Session imported successfully

Chat with HAR

Explore flows, auth, retries, and edge cases.

Selected

Replay in sandbox

Re-run any captured request and see the real response.

Debug with HAR

Trace failures and inspect request-response behavior.

HAR

Chat with HAR

148 requests
Extract the login flow, include token refresh, and point out any required headers.

Assistant

`POST /auth/login` sets the session cookie.
`POST /auth/refresh` rotates the bearer token.
Requires `x-device-id` and `x-region` headers.
Give me the API to download all the photos
arrow_upward

Use Cases

Your HAR file knows everything. Ask it anything.

Most tools hit a wall when your HAR grows large. TraceMiner is purpose-built for it — thousands of requests, megabytes of trace data, processed without truncation.

api Reverse Engineer Zero source code needed

X-ray any website's backend

Curious how your favourite app actually works? Upload its HAR and ask TraceMiner anything — which APIs get called, what each returns, how auth flows, how data is paginated. The entire backend, laid bare.

/api/products
78%
/api/auth
52%
/api/media
90%
data_object Data Scraping Sandbox replay

Scrape data — zero code written

TraceMiner infers API patterns from HAR responses and replays them in a sandboxed environment. Just say what you need — download the results or export a ready-to-run cURL or Python script.

CSV PDF JSON Word cURL Python
compare_arrows Comparison HAR reality check

Why generic LLMs fail

memory

Big HAR files break them

Generic LLMs hit context limits fast when the trace gets large

block

Important requests get lost

Long chains are dropped, compressed, or only partially read

psychology_alt

They start making things up

Once context runs thin, answers can sound sure but be wrong

vs
diamond
TraceMiner Grounded
all_inclusive

Handles full HAR size

Upload huge HAR files and TraceMiner keeps working without sweating

target

Pulls the exact signal

Finds the real flows, endpoints, auth steps, and failures inside the noise

verified

Tells you when data is missing

If the trace did not capture something, it says so clearly instead of hallucinating

troubleshoot Debug & Document

Debug issues & generate docs instantly

Drop the HAR, ask what broke, and TraceMiner surfaces the failing request, explains the root cause, and turns the flow into polished docs your team can use immediately.

bug_report Debug description Docs account_tree Flow bolt Fixes

Upload the working HAR. Upload the broken one. Find what broke fast.

A simpler flow: compare both traces, let TraceMiner isolate the split, and get a clear explanation of the request or state change that caused the failure.

Working vs failing

Trace comparison

Compare the working flow with the failing one.

Drop in both HAR files and TraceMiner lines up the real request chain so you can stop guessing where the break happened.

search_insights

Root cause signal

Mismatch surfaced

Surface the exact mismatch that caused the bug.

Headers, cookies, redirects, anti-bot checks, payload changes, missing steps. The trace tells the story once the noise is removed.

bolt

Actionable output

Short path to fix

Keep the output simple: what changed, where it changed, and why it failed.

Instead of scrubbing thousands of requests by hand, you get a short path to the root cause and a concrete explanation you can act on immediately.

Working HAR

200 on callback

Cookie and redirect chain stay intact.

Non-working HAR

302 loop on callback

`state` cookie drops before the OAuth callback returns.

TraceMiner diff

mismatch isolated
Flow diverges after `GET /oauth/callback`.
Working trace sends `cookie: state=...` before redirect follow-up.
Broken trace loses the cookie, so auth restarts and loops.
bolt For AI Agents

Every website has a private API. Now your agent can use it.

Browser automation makes an agent read a page like a human — slowly, expensively, and it breaks on every redesign. TraceMiner gives it the API underneath instead: the same calls the site makes, with auth already worked out.

01
manage_search
AI Discovery

Surface hidden APIs in seconds

Private endpoints that take months of manual reverse engineering are exposed in a few clicks. AI agents sweep your entire HAR — no matter the size — and filter exactly what you asked for. Payment flows, product catalogues, auth tokens — just tell the AI and it zeroes in.

02
key
Auth Mapping

Auth, headers, and session state decoded

The hard part of driving a site programmatically was never the request — it was the token refresh, the signed header, the cookie that expires mid-flow. TraceMiner reads the real chain out of the trace and tells you exactly what a call needs to succeed.

03
robot_2
Sandbox Replay

Replay it live until it works

Run any request in an isolated sandbox before you build on it. Wrong headers? Expired cookie? Unexpected schema? You see the real response, not a guess — so the flow you hand your agent is one you have already watched return 200.

04
rocket_launch
Agent Automation

Hand it to your agent. Sky's the limit.

Plug the verified flows into Claude Code or any agent runtime and let it work the site directly — no headless browser, no brittle selectors, no per-page compute bill. Direct API calls are faster, cheaper, and they do not break the next time someone ships a CSS change.

electric_bolt

What used to take an entire engineering sprint now takes minutes.

No API docs. No insider access. No months of manual tracing. Capture the session once and your agents work the site through its own API.

Start with a HAR arrow_forward Book a Demo

Built for people tired of hitting locked doors on the web.

search

Personal data visibility

Answer the questions closed products never answer cleanly, like how much you spent, what changed in your account, or which requests power the dashboard you already use.

fact_check

Agentic workflow automation

Capture a workflow once, expose the real request chain, and turn repetitive admin work into something your agent can actually execute.

sync_alt

Product teardown and debugging

Understand how a product really works, compare working and broken HARs, and isolate the exact request, cookie, or header that changed.

127+
Requests analyzed inside a typical HAR
4
Minutes to a usable understanding of the flow
Grounded
Answers from captured traffic, not hallucinated guesses

Pay as you go

Simple credits. No subscription.

Start from $2, top up whenever you need, and use credits at your own pace over the next 6 months.

Starter

5k credits

First-time users only
$2 $5 one time

~ 500K tokens · 1 credit = 100 tokens

Test the full workflow and see if TraceMiner earns a place in your stack.

Buy for $2

Builder

10k credits

Popular
$10 one time

~ 1M tokens · 1 credit = 100 tokens

A practical pack for regular analysis, debugging, and sandbox replay.

Buy for $10

Operator

30k credits

Best value
$25 one time

~ 3M tokens · 1 credit = 100 tokens

For heavier usage across repeated runs, multiple targets, and deeper iteration.

Buy for $25

Ready to Start

Stop asking for access.
Upload the HAR and start building.

The web was already speaking through its network layer. Capture the trace, let TraceMiner read it, and ship the thing that used to die behind closed APIs.

Start using TraceMiner
Do I need to format the HAR in a special way? +

No. Export the HAR from your browser session and upload it as-is. TraceMiner reads the captured traffic and lets you ask for the flow, bug, or endpoint you care about.

What if the HAR is large or noisy? +

That is exactly the problem TraceMiner is built for. It pulls signal out of thousands of requests and tells you clearly when something was never captured.

Can I use this to drive a site from an AI agent? +

That is the main use case. TraceMiner gives you the real request chain — endpoints, headers, and auth — behind a flow, and you can replay it in the sandbox before handing it to Claude Code or any agent runtime. Direct API calls instead of a headless browser.