# BlazeMeter MCP for AI Agents MCP > The BlazeMeter MCP automates continuous performance testing by letting your AI agent manage cloud load tests, workspaces, and metrics directly through chat. It lets you trigger stress tests, analyze throughput KPIs like p90/p99, and safely shut down runaway connections—all without switching dashboards. ## Overview - **Category:** ship-it - **Price:** Free - **Tags:** performance-testing, load-testing, stress-testing, continuous-testing, qa-automation, infrastructure-monitoring ## Description Managing large-scale performance tests used to mean context switching: jumping between your CI/CD pipeline, the testing console, and a metrics dashboard. Now, you can keep everything in one place. This MCP connects BlazeMeter’s full suite of enterprise load testing capabilities directly to any AI client. Your agent handles everything from listing available workspaces and projects to executing complex load tests against your target APIs. It reads live run data, giving you critical throughput reports (p90/p99 KPIs) instantly. Need to stop a test immediately? You can force shut down runaway connections with a single command. If you're already using the Vinkius catalog for other services, adding BlazeMeter centralizes your entire DevOps toolchain under one roof. It lets SREs run rapid baseline tests and QA teams verify JMeter structures, all while staying in their natural language environment. ## Tools ### list_workspaces Identifies and lists all bounded workspaces within the BlazeMeter platform. ### list_projects Extracts a list of projects that are bound to a specific workspace. ### list_tests Provides a comprehensive JSON payload listing all available tests. ### get_test Retrieves the full configuration for an active test limit within the vault. ### start_test Initiates a new, irreversible load generation validation run using specified metrics. ### list_masters Enumerates all attached structured rules that export active master records. ### get_master Runs an automated validation check to determine the status of a specific gateway run. ### stop_master Sends a native Gateway shutdown logic command to identify and stop active master arrays. ### get_report Inspects deep internal data arrays to mitigate specific plan math reports. ### get_user Identifies and retrieves the active user records associated with the platform. ## Prompt Examples **Prompt:** ``` I need to run a stress test on my main user API. Can you start it? ``` **Response:** ``` **Performance Test Status** * **Test ID:** 987654 (UserAuth) * **Status:** Running (Active Master: m-11223) * **Throughput:** Stable, currently at 120 RPS. I've initiated the load test. I recommend monitoring the p99 KPI for spikes over the next hour. ``` **Prompt:** ``` List all my active workspaces and projects so I know what tests are available. ``` **Response:** ``` Here is a list of your current testing resources: **Workspaces Found:** * `Development_API` (ID: 123456) * `Pre-Production_Staging` (ID: 789012) Under `Development_API`, the following projects are configured: * `Core Backend Stress` * `User Profile Load Test` ``` **Prompt:** ``` Stop that master test right now. It's running too long. ``` **Response:** ``` ⚠️ **EMERGENCY SHUTDOWN INITIATED** ⚠️ Connection to Master ID `m-11223` has been successfully severed and scaled down. All active load generation for this session is terminated immediately. You can now review the final metrics report in your chat history. ``` ## Capabilities ### Inventorying Testing Resources List all available workspaces, projects, and structural user metadata within the BlazeMeter platform. ### Executing Load Tests Start cloud-based performance tests using configured JMeter definitions to simulate real-world load on your system. ### Monitoring Live Test Runs Query the operational health of active master runs and retrieve precise throughput reports, including p90 and p99 metrics. ### Managing Master Connections Enumerate attached structured rules and check the status of gateway run validations for critical systems. ### Stopping Active Tests Forcefully shut down active cloud connections or runaway master runs to protect your source architecture during testing. ## Use Cases ### A critical API endpoint is slowing down during peak usage. The SRE asks the agent to run a stress test on the payment gateway. The agent uses `start_test` and then queries the resulting data with `get_report`, immediately showing that p99 latency spikes above acceptable limits, identifying the bottleneck for the team. ### Need to quickly validate a new microservice deployment. The QA Automation Team uses the agent to list all available test structures (`list_tests`). They select the correct definition and trigger an execution. The system confirms the run is active, allowing them to proceed with confidence in the CI/CD cycle. ### A performance test run gets out of control. During a large-scale deployment simulation, the connection runs wild and threatens stability. Instead of scrambling for the dashboard, the DevOps Engineer tells the agent to use `stop_master`, safely isolating the network and preventing an outage. ### Need to audit who has access to test environments. A manager needs a full list of users with testing permissions. The agent uses `get_user` to pull active user records, allowing them to quickly verify the roles and account metadata across the platform. ## Benefits - Stop switching tabs. Instead of jumping between dashboards to check load results, your agent manages everything from workspaces (using `list_workspaces`) to live run monitoring. - Gain granular control over test environments. Use the MCP to list projects (`list_projects`) and retrieve detailed configurations for specific tests (`get_test`) before running them. - Respond instantly to failures. If a test runs wild, you can execute emergency shutdown controls using `stop_master` without manual intervention or complex API calls. - Get immediate visibility into performance health. Querying active master records via `list_masters` and checking gateway run status with `get_master` provides instant operational insights. - Eliminate guesswork during testing. You can trigger a new execution (`start_test`) and immediately query the resulting report data using `get_report`, getting p90/p99 KPIs right in your chat. ## How It Works The bottom line is that your AI agent translates high-level instructions into precise API calls, giving you immediate control over your entire performance testing lifecycle. 1. Subscribe to this MCP and input your BlazeMeter Key ID and Secret credentials. 2. Connect your preferred AI client (Claude, Cursor, etc.) to the Vinkius catalog using these stored credentials. 3. Issue natural language commands through your agent, allowing it to execute complex load generation validations directly against the platform. ## Frequently Asked Questions **How do I use BlazeMeter MCP to test my API performance?** You tell the agent which endpoint you want to stress-test. The MCP triggers a cloud load run, simulating real user traffic. You get immediate access to throughput reports and latency metrics like p90/p99 right in your chat window. **Can I use BlazeMeter MCP if my test runs go wrong?** Yes. If a test gets out of control, the agent can execute emergency stop controls using natural language. This safely shuts down runaway connections and protects your network architecture instantly. **Does BlazeMeter MCP handle multiple environments?** Absolutely. You can list all available workspaces and projects first. Then, you direct the agent to run a specific test on any environment—from development to pre-production—without switching context. **What metrics does BlazeMeter MCP give me?** You get precise operational health data. This includes throughput reports and critical KPIs like p90 (the 90th percentile) and p99 (the 99th percentile), which tell you how the system performs under heavy load. **Is BlazeMeter MCP better than using a standalone dashboard?** Yes. It keeps your entire testing process—from listing resources to executing tests—in one place with your AI client. You don't need to leave your conversation or IDE.