Load test any API
right in your browser.
Hit fires thousands of requests at your endpoints. REST or streaming AI, doesn't matter. You see latency, throughput, time-to-first-token, and cost as it runs. No account. No backend. Nothing leaves your machine.
Built for the people who break things first,
API developers · QA engineers · SREs · AI builders
A real load tester. Lives in a popup.
Runs locally in your browser. Use the popup for quick hits. Open Studio for full runs.
LLM load testing New
k6 and Locust miss the metrics that matter for streaming AI: time-to-first-token, tokens/sec, cost per request. Point Hit at any SSE or chat/completions endpoint. You'll know in seconds if the UX holds up.
Live charts
Latency with p90 and p99 bands, throughput, status mix, histogram. Updates as the run goes.
Assertions
Status code, latency budget, body contains, JSON path, max TTFT. The run ends with a pass or a fail.
cURL import
Copy a request as cURL from DevTools, paste it in. Method, headers, body. All set up.
Load patterns
Burst, constant, ramp-up, spike. Match real traffic. Find the breaking point.
Real payloads
Unique data per request with {uuid}, {int}, {email}, ${VAR}. Stop hitting the cache with static junk.
Up and running in 30 seconds
No store account, no build step. Loads as an unpacked extension. (Chrome Web Store listing coming soon.)
Load it in Chrome
Open chrome://extensions, turn on Developer mode, click Load unpacked, and pick the unzipped folder.
edge://extensions
Pin it & hit Run
Click the Hit icon, add an API (or paste a cURL), and press Run. Open Studio for charts and AI metrics.
The only browser tool that load tests streaming AI
Other tools give you one number: total response time. That number hides the thing your users actually feel. A test can pass while users wait 8 seconds for the first word.
Hit measures what matters for AI:
- TTFT: time to the first token (p50, p90, p99)
- Tokens/sec: the speed users feel
- Cost per request: based on a price you set
- Assertions on
maxTTFTandminTPS
Why reach for Hit
Postman
Solid client. Bad at load. The free tier keeps shrinking.
k6 / Locust
You write scripts. They measure streaming AI wrong.
Cloud load testers
Signup, per-request fees, your data on someone else's box.
Hit
30 seconds to install. No config. Knows AI metrics. Free. All local.
Good to know
Is it really free?
Yes. Hit is free. It runs in your browser. When you need continuous, team-wide monitoring, the alt.qa platform takes over.
Where does my data go?
Nowhere. Hit has no backend. Requests go from your browser straight to the endpoints you point it at. Config is stored locally.
Why isn't it on the Chrome Web Store yet?
It will be soon. For now you download, unzip, and load it as an unpacked extension. The whole thing takes 30 seconds.
Can it test endpoints behind login or CORS?
Yes. It's a browser extension with host permissions, so CORS doesn't block it and it reuses your existing session. No token copying.
How does the AI testing work?
Turn on Streaming for an API. Hit reads the response token by token and measures TTFT, inter-token gaps, tokens/sec, and cost. Works with OpenAI and Anthropic SSE formats out of the box, with a generic fallback for others.
Will it crash production?
It's a load tool, so point it at staging. If you must hit prod, keep bursts and concurrency low. Every request includes X-Hit-Test: true by default. You can stop a run instantly.
Start hitting your APIs.
Free. Local. 30 seconds.
Add to ChromeVerify your download
9c1d0fd8ef5505e570d3bf9e6a4de37a88ca80a01ab8300af9dc757d443c3a9bd790d4b7debfd8e4e88a0791b4b99cfa