Slides are shown one under the other. Presentation mode needs JavaScript. All presentations · Türkçe sürüm
Testnizer Product overview · v1.6.0

The offline API and AI client that also does your security work.

Test APIs, MCP servers and LLMs, mock all of them locally, and do the cryptography on your own machine — free, no account, no telemetry.

HTTPSOAPGraphQLgRPCWebSocketSocket.IOMQTTSSEMCPAI Chat
macOS · Windows · Linux

What it is

One desktop app for the whole API workbench.

Testnizer is a free desktop client for macOS, Windows and Linux. It runs entirely on your computer: nothing to sign up for, no server to install, no cloud in between.

Test

APIs, MCP servers and LLMs

REST, SOAP, GraphQL, gRPC and real-time protocols, an MCP client, and AI Chat across 14 providers — with scripts, tests and the Runner.

Mock

Every side of the conversation

HTTP, SSE, AI (OpenAI and Anthropic APIs) and MCP mock servers on 127.0.0.1, with the latency and failures you choose.

Secure

The cryptography, on your machine

Keystores, mTLS, JWT / JOSE, JWK, SAML, WS-Security, a TLS inspector and an MCP security scan — the private key never leaves the app.

Built for regulated teams — banking, government, insurance and healthcare, where production credentials must never leave the machine.

And for every developer — free, no sign-up, and your Postman collections and scripts come with you.

Why offline

Requests go to your targets. Nothing goes to us.

No account, no cloud login, no sync
No telemetry — there is no analytics endpoint
No background traffic from the interface

The interface cannot open a network connection of its own: every request runs in the app core and goes only where you send it. Workspaces, history, certificates and secrets live in one local database. The only request of its own is the update check, which you can turn off — air-gapped networks are supported.

One client, every protocol

Eight API protocols, plus MCP and AI Chat.

Each one has its own editor, not a generic HTTP form — and they all share the project's environments and {{variables}} and save with Ctrl/Cmd S.

HTTP / REST

Every body mode, path variables, cookies, mTLS, proxy; paste a cURL into the URL bar

SOAP / WSDL

WSDL import, SOAP 1.1 + 1.2, WS-Security on Send and in the Runner

GraphQL

Queries, mutations, schema explorer, subscriptions over WebSocket

gRPC

.proto import, all four streaming modes, metadata

WebSocket

Text, JSON, XML, Base64, hex; saved messages, searchable log

Socket.IO

Client v2, v3 or v4, namespaces, acknowledgements

MQTT

MQTT 3.1.1 and 5.0 — QoS, retain, last will, TLS

Server-Sent Events

Long-lived streams, custom headers, Last-Event-ID resume

MCP

Streamable HTTP, SSE and stdio; tools, resources, prompts

AI Chat

14 providers plus a custom URL, streaming, MCP tools

AI & MCP · the client

Everything an MCP server offers, in one tab.

  • Streamable HTTP, legacy SSE or stdio — for stdio, Testnizer starts the local server
  • Protocol 2024-10-07 to 2026-07-28: Auto, Legacy or a pinned version
  • Tool arguments as a form built from the JSON Schema; resources, prompts, notifications
  • The raw JSON-RPC frames, handshake included
  • An OAuth 2.1 debugger: metadata, client registration, PKCE, sign-in and token exchange, step by step — tokens stay in the app core
  • Save calls with tests and scripts (pm.mcp) and run them in the Runner, a Test Suite or a schedule

AI & MCP · security scan

An A–F report on an MCP server, every finding with its evidence.

30checks in eight categories
+6for servers that speak 2026‑07‑28

How the grade is computed

The score starts at 100. A failed check subtracts its severity weight (critical 25, high 15, medium 8, low 3); a warning subtracts half.

A≥ 90B≥ 80C≥ 65D≥ 50F< 50
Runs from your computer against the server you point it at; no AI model is involved and credentials are scrubbed from every finding. The report is one self-contained HTML file. Scan only servers you are authorized to test.

AI & MCP · AI Chat

A chat client that calls your MCP tools — after you say yes.

  • 14 providers plus a custom URL — OpenAI-compatible or Anthropic; Ollama and LM Studio work too
  • Every tool call asks first by default. “Run tools without asking” lasts for the open tab only and is never saved
  • Each reply shows status, time to first token, duration and token usage — “not reported” when the provider says nothing, never 0
  • API keys are encrypted per provider with the operating system's key store
  • Conversations stay on this computer, masked when written — never exported or committed

Mock servers

Mock the API, the stream, the model and the tools.

Real servers inside the app, on 127.0.0.1 by default. Build and test a client or an agent without a provider key, without spending tokens and without a network.

HTTP

Rule-based endpoints

Conditions, templating, scripts, auth, schema validation, failure injection, rate limits, a recording proxy.

SSE

An event stream you define

Interval, heartbeat, loop, Last-Event-ID resume, and an error after N events.

AI

OpenAI and Anthropic APIs

Streaming, token usage, tool-call round trips, 429 / 500 / 401, Anthropic 529 and failures mid-stream.

MCP

A local MCP server

Tools with JSON Schema, resources, prompts, bearer auth, latency, error scenarios; 2025 and 2026-07-28 clients.

Automation

Run a whole flow, clean up after it, keep the report.

Setup→Flow→Teardown always runs
  • Iterations and data files (CSV, JSON), with variables chained from one request to the next
  • A critical request stops the run when it fails; the report names it
  • HTML and JSON reports — one self-contained file, credentials masked
  • Test Suites span several collections and run module by module
  • Scheduled runs: interval, daily, weekly or cron while the app is open; a Run fail script can post the report to a webhook

Bring your work with you

Any Postman export comes in. Your project goes back out.

Import — recognised from the file
Postman v1v2.0v2.1v3 JSONv3 Native Git folderEnvironmentGlobalsData dump.zip
Export
Postman v2.1Postman v3OpenAPI 3InsomniaTestnizer JSON
pm.*one script runtime — Send and Run behave the same
118Postman dynamic variables, {{$randomEmail}} …
cURLpaste into the URL bar — bash, Windows CMD, PowerShell
36Generate Code variants

Also imports OpenAPI 3.x, Swagger 2.0, Insomnia v4 / v5, WSDL, .proto, RAML 1.0, SoapUI / ReadyAPI and HAR. Every import ends with a report of what was converted, skipped or could not be carried — Kafka and Postman Data requests in a v3 collection are listed as skipped.

Security toolbox

The crypto work you used to do in a terminal — or on a website.

Keystore Studio

JKS and PKCS#12 without keytool: open, create, convert, inspect certificates

JWT / JOSE

Decode, sign, verify, encrypt and decrypt — also as pm.jose in scripts

JWK / JWKS

PEM ⇄ JWK, RFC 7638 thumbprints, a JWKS served by the mock server

SAML 2.0

Build, sign and validate — signature wrapping is rejected

WS-Security

UsernameToken, Timestamp, XML signature and encryption on the SOAP request

TLS Inspector

What a live endpoint presents: chain, protocol, ciphers

mTLS & certificates

Client certificates per host, from PEM, PFX or a keystore entry

Passwords, OTP, QR

Plus JSONPath, XPath, XSLT, Jolt, diff, hashes and encoders

Everything runs in the app on your computer. A saved keystore supplies the key for mTLS, WS-Security, JWT and SAML — the private key never leaves the app.

Teamwork

Your collections in your Git repo, reviewed like code.

  • Each project maps to a branch; switch branches without a terminal
  • Commit…, Pull, Push, Merge from the branch menu, with ahead / behind counts
  • A teammate's pull brings the whole project: project auth, scripts and settings
  • Secret values are written empty in every committed file; a pull says how many are still empty on this machine
  • Scripts in a cloned project do not run until you allow them on this computer

GitHub, GitLab, Bitbucket, Gitea or any Git server over HTTPS or SSH.

Privacy

Your secrets stay on this machine — on disk, in files and on screen.

At rest — encrypted

Secret variables, mock credentials, certificate passphrases, Git tokens and AI keys are encrypted with the OS keychain (Keychain, DPAPI, libsecret). Where there is none — some Linux setups — they stay unencrypted and the app says so.

In files — left out

Secret values are written empty in every Git commit, project file and export, and a pull or import never overwrites your local value.

On screen and in logs — masked

Credential headers, fields and secret values are masked in History, the Console, run reports and mock logs before anything is stored or shown.

No telemetry No account Optional app password — a lock on launch, not database encryption A database backup when a new version first starts

Platforms

macOS, Windows and Linux. Free.

macOS

Apple Silicon and Intel. Signed with an Apple Developer ID and notarized. macOS 13 Ventura or later.

arm64x64

Windows

x64 and arm64. The installer is code-signed.

x64arm64

Linux

Debian / Ubuntu packages and AppImage, x64 and arm64.

debAppImagex64arm64
Download Testnizer No account, no subscription. The source code is on GitHub under the MIT License.
Testnizer

The offline API and AI client that also does your security work.