Goku

What is Goku?

Goku is a high-performance, scalable HTTP load-testing tool designed for benchmarking and performance analysis of web services. It includes an MCP (Model Context Protocol) server that enables LLMs and AI agents to programmatically run load tests and retrieve structured results.

How to use Goku?

Install the MCP server via Cargo (cargo install goku-mcp) or the provided install script. Once registered as an MCP tool, you can ask an LLM to run a benchmark using the run_benchmark tool with parameters such as target URL, number of concurrent clients, test duration, and more.

Key features of Goku

• Real-time live stats during long tests
• Rate limiting (--rps) for constant-rate load profiles
• Multi-step sequential scenarios (multiple endpoints per test)
• Variable templating in URLs and bodies
• Built-in authentication (Bearer token, Basic auth)
• MCP server for LLM/agent integration

Use cases of Goku

• Ask an LLM to run a load test on an endpoint and return p95/p99 latency
• Automate performance regression testing in CI/CD pipelines via an agent
• Compare two benchmark runs to detect performance changes
• Simulate gradual traffic spikes with ramp-up and rate limiting

FAQ from Goku

What is the Goku MCP server?

It is a Model Context Protocol server that exposes the run_benchmark tool, allowing LLM agents to perform HTTP load testing without manual CLI usage.

How do I install the Goku MCP server?

Install via Cargo (cargo install goku-mcp) or run the automatic install script provided in the repository.

What parameters does the run_benchmark tool accept?

It accepts target URL, number of clients, total requests or duration, request body, headers, HTTP/2 flag, ramp-up, timeout, insecure TLS, and rate limit (RPS).

Can I use authentication in my load tests?

Yes, Goku supports Bearer token and Basic authentication via dedicated flags in the CLI, and similarly through the MCP tool’s parameters.

How do I compare two benchmark runs?

Use the compare subcommand with two JSON result files to see a colored diff table of performance metrics.