Load test your app from your coding agent
Ask Claude Code, Cursor or any MCP client to load test your app before you launch. The agent already knows your code, so it picks the port and the pages worth testing. Stresix builds the app on a fresh cloud server, ramps simulated users at it from a second server, and gives the agent the report: what ran out first, p95, error rate and what server size to run it on.
Set it up in two minutes
Sign in at stresix.com with GitHub, Google or email. Open Settings, go to Keys for agents and create a key. The key is shown once, together with the command above already filled in.
Add the server to your agent with the command or the JSON above. Any client that speaks MCP over HTTP works the same way: the server is https://stresix.com/mcp and the key goes in the Authorization header.
Then ask in plain words, for example: load test this app before I launch, and tell me if a 2 vCPU server is enough.
What the agent can do
start_load_test takes a GitHub repository URL or a Docker image, the port, up to 10 GET paths, a branch and the run size. It saves the run and starts it.
get_test_status says where a run stands: building, starting servers, or running with live users, requests per second, p95 and error rate.
get_test_report returns the verdict when the run is done: how many users it held, p95, error rate, what ran out first, what to change and the server size to pick, plus a link to the full report with charts.
That is all a key can do. It cannot buy credits, change your account or see other people's runs.
What it costs
Your first run is free with a GitHub account at least a week old: Lite sends up to 500 users for 4 minutes at a 2 vCPU / 4 GB server. The agent uses it first when you do not name a size.
After that, Standard (2,500 users, 10 minutes, 4 vCPU / 8 GB) costs 1 credit and Massive (10,000 users, 15 minutes, 8 vCPU / 16 GB) costs 3. Credits come in packs, $10 for 5 or $40 for 25.
A paid run never starts on its own. The tool first answers with the price and your balance, and the agent has to ask you before it calls it again with confirm_spend. With no credits left the tool says to top up at stresix.com, and nothing is started.
Guard rails for agents that loop
Agents retry. Each key can start 4 runs an hour, on top of the account's daily limit, so a stuck agent cannot spend a balance. A run that fails on the Stresix side is refunded.
Revoke a key in Settings at any time and it stops working at once. Only a hash of each key is stored.
claude mcp add --transport http stresix https://stresix.com/mcp \
--header "Authorization: Bearer YOUR_STRESIX_KEY"
{
"mcpServers": {
"stresix": {
"url": "https://stresix.com/mcp",
"headers": { "Authorization": "Bearer YOUR_STRESIX_KEY" }
}
}
}Frequently asked questions
Which agents does it work with?
Clients that can add a remote HTTP server with a custom header: Claude Code, Cursor, Windsurf, VS Code and others. Claude Desktop custom connectors expect an OAuth sign in, which the server does not offer yet.
Does the agent need my GitHub token?
No. Public repositories work as they are. For a private repository, connect GitHub once in Stresix Settings; the agent never sees that token.
Can the agent test any website?
No. Stresix only tests the app it builds from your repository or image, on its own server. It never sends load at a live site.
Is there something to install?
No. The server runs at stresix.com/mcp, so there is no package to install or keep updated.
Related guides
Stop guessing your specs. Start knowing them.
Point Stresix at a repo or Docker image and get a number you can deploy on, in one run.
Start a load test