The whole lifecycle is covered — upload the script, set concurrency and ramp-up, choose locations, start it, then read the summary, the errors and the request statistics — so "change one thing and rerun" stops meaning a browser tab and a wait. The cost estimate is the part that earns its keep on a credit plan: you can price a configuration before it consumes anything.
BlazeMeter's own server for its cloud performance-testing platform. It covers the whole hierarchy — account, workspace, project, test, execution — so a load test can be configured, started, watched and reported on without leaving the conversation.
- Tests created and configured: users, duration and ramp-up, geographic distribution, and uploaded scripts and assets
- The catalogue of pass/fail rules read, then enabled or disabled on a test
- Tests started, execution status followed, and every execution for a test listed
- Reports pulled back individually or all at once: summary, error analysis, request statistics, anomaly statistics
- AI analysis of a test launched and read back
- A cost estimate for a configured run against your allowance type, before the credits are spent
- Your own profile with the default account, workspace and project IDs — where everything else starts
- Accounts, workspaces and projects listed and inspected, with available test locations, billing usage and test counts
- BlazeMeter help categories, articles and curated skills read in place, so product questions get answered from official content
BlazeMeter API credentials — an API Key ID and secret, downloaded as api-keys.json. The prebuilt binary reads that file's path from BLAZEMETER_API_KEY and is launched with --mcp; run it once and it prints a ready-made config plus install links for Cursor and VS Code. From source instead, uvx with Python 3.11 or newer. The ghcr.io/blazemeter/bzm-mcp Docker image takes API_KEY_ID and API_KEY_SECRET directly, with SOURCE_WORKING_DIRECTORY bound to your test files. AI consent has to be enabled on the account by its owner. Tool-call traces and metrics are exported by default — OTEL_SDK_DISABLED=true or --no-telemetry stops that, OTEL_EXPORTER_OTLP_ENDPOINT redirects it to your own collector. Behind SSL inspection, point SSL_CERT_FILE at your mounted CA bundle.