NewNative agents for Java, .NET, Python and Go

See what flows
between your systems

Your CRM talks to your support desk, which talks to core banking. TraceTap maps those connections, learns the shape of the data on each one, and tells you the moment something diverges.

Nothing sensitive reaches us. Sampled traffic is redacted inside the probe, on your own machine — and a field nobody has classified is masked by default.

POST /api/customersemail: ma••••••••leiban: NL•••••••••00amount: 1250.50CRMSalesforceSupport deskZendeskPaymentsAdyenIntegration busNode.jsCore bankingSOAP bridgeWarehouseSnowflake

Taps the systems you already run

TomcatJBoss / WildFlyWebSphereSpring BootASP.NET CoreDjangoFastAPIFlaskExpressNext.jsGoAnything over HTTP
The invariant

We never see your customers' data

Every other tool in this category asks you to ship production payloads to a vendor. TraceTap does the opposite: the probe transforms the sample on your own hardware, before it goes anywhere.

The default for a field nobody has classified is to mask it. A body the probe cannot parse is withheld entirely — because a body it could not inspect is a body it could not redact.

Redacted at the source

Rules run in the probe, on your machine. Not in our cloud.

Encrypted end to end

AES-256-GCM under a key only your project holds.

Realistic, not blank

Deterministic fakes: valid IBANs, Luhn-safe cards, stable across systems.

A human approves

AI can suggest. Nothing is published until a person signs it off.

Four steps to a landscape that watches itself

From an empty canvas to rules your probes enforce, without writing a schema by hand.

  1. 01

    Map the landscape

    Draw every system and the links between them on a canvas. Say what each link speaks — JSON, XML, SOAP, gRPC — and how you can tap it.

  2. 02

    Install a probe

    A proxy binary, a servlet filter, ASP.NET Core middleware, WSGI, Go — whatever fits. One in every N messages is sampled; the rest are only counted.

  3. 03

    Watch it learn

    TraceTap builds a field-by-field picture of every endpoint. Anything that does not match what it learned becomes a divergence in the Pulse feed.

  4. 04

    Approve, then publish

    Say what each field is, in plain language, with a live before-and-after. Approving publishes rules the probes enforce locally from then on.

Everything, from the map to the payload

Screenshots from the real product, not a mock-up.

Topology

Draw the landscape once

Every system, every link, and how each one is tapped. Drag a box, connect two systems, pick the format on the wire. The map is what everything else hangs off.

  • JSON, XML, SOAP and gRPC on any link
  • Marks which links have no probe reporting
  • Saves the whole canvas in one action
The topology editor showing a CRM, support desk, integration bus, core banking and payment gateway connected by labelled links

Review

Decide what every field is — and see it happen

A non-technical reviewer works down a list in plain language. The before-and-after re-renders as they go, so "is this personal data?" is answered by looking, not by guessing.

  • Built-in heuristics flag the obvious cases
  • Ask an AI to classify a batch — you still approve
  • Custom generators in JavaScript, with a live preview
The review workspace listing learned fields with classifications, actions and generators

Statistics

Real traffic volume, not a sample count

Probes count every message they see, not just the ones they keep — so the numbers are the traffic, not the sampling rate. Estimates are labelled as estimates.

  • Measured throughput against what was captured
  • Per connection, per system, per probe, per endpoint
  • Latency, error rate and response mix
The statistics screen with throughput over time, a sampling meter and per-connection breakdowns

Pulse

Told the moment something changes

A new field, a changed type, a field that vanished, a 5xx that never used to happen. Each one arrives once, with the evidence attached — and email that never becomes a firehose.

  • One item per change, not one per sample
  • AI can explain a divergence and suggest a fix
  • Resolve, ignore, or jump straight to the field
The Pulse feed showing project health cards and a list of divergences awaiting a decision
Ten ways in

A probe for the stack you actually have

Including the ones nobody else supports. If the system is twenty years old and you cannot change a line of it, put the proxy in front and point the client at that.

Not one agent carries a third-party dependency — no Jackson in your WAR, no packages in your virtualenv, nothing new in your go.mod.

Read the API docs

Java servlet filter

Tomcat · JBoss EAP · WildFly · Jetty · WebSphere · Spring Boot

Java 8+, one JAR, zero dependencies

ASP.NET Core

Kestrel · IIS in-process and out-of-process

One line in Program.cs

Python WSGI / ASGI

Django · Flask · FastAPI · Starlette · Pyramid

Wrap the application object

Go net/http

stdlib mux · chi · gorilla/mux · echo · gin

Wraps any http.Handler

Proxy binary

Anything you cannot change

Sits in between, passes bytes through

Node libraries

Express · Next.js · fetch() · axios

Two lines, no redirect needed

Find out what your systems are really saying to each other

Map a landscape, install one probe, and watch the first divergence arrive. Your data never leaves your network in the clear.

TraceTap — see what flows between your systems