MCP Drill

A high-performance stress testing platform for MCP servers and gateways.

Simulate thousands of concurrent MCP clients, identify performance bottlenecks, and validate your infrastructure before production.

Built to battle-test Peta, an MCP control plane and runtime. If you're building MCP infrastructure, MCP Drill helps you break it before your users do.

Why MCP Drill?

• Self-Contained - Run everything on your own infrastructure. No external services, no cloud dependencies
• Realistic Load Simulation - Configurable operation mixes with weighted tool selection
• Multi-stage Testing - Preflight, baseline, ramp-up, and soak stages
• Real-time Observability - Live metrics dashboard with SSE streaming
• Distributed - Scale horizontally with multiple workers
• Safety First - Built-in stop conditions prevent runaway tests

Architecture

┌──────────────────────────────────────────────────┐
│                    Control Plane                 │
│  ┌────────────┐  ┌────────────┐  ┌────────────┐  │
│  │Run Manager │  │ Scheduler  │  │  HTTP API  │  │
│  └────────────┘  └────────────┘  └────────────┘  │
└─────────────────────────┬────────────────────────┘
                          │
            ┌─────────────┼─────────────┐
            │             │             │
      ┌─────┴─────┐ ┌─────┴─────┐ ┌─────┴─────┐
      │  Worker   │ │  Worker   │ │  Worker   │
      │ (VU Pool) │ │ (VU Pool) │ │ (VU Pool) │
      └─────┬─────┘ └─────┬─────┘ └─────┬─────┘
            │             │             │
            └─────────────┼─────────────┘
                          │
                   ┌──────┴──────┐
                   │ MCP Target  │
                   └─────────────┘

Screenshots

Real-time performance monitoring with throughput, latency percentiles, error rates, and server resource utilization:

Per-tool breakdown with success rates, latency distribution, and error analysis:

Quick Start

Option A: Docker (Recommended)

Prerequisites: Docker Engine + Docker Compose v2 (docker compose).
If you see build warnings, also ensure Buildx is installed (docker buildx version).

git clone https://github.com/bc-dunia/mcpdrill.git
cd mcpdrill

# Core (server + worker)
docker compose up -d

# With mock server for testing
docker compose --profile mock up -d

# Full stack (mock + agent)
docker compose --profile full up -d

Open http://localhost:8080/ui/logs/ — the Web UI is embedded in the server.

Scale workers: docker compose up -d --scale worker=5

Notes

• Ports are bound to localhost by default (see docker-compose.yml). Update the port mappings if you need LAN/remote access.
• The default compose config runs in insecure mode for local development (--insecure, --insecure-worker-auth) and enables private-network discovery (--allow-private-discovery) for Docker testing. Don’t expose it publicly without enabling auth.
• API rate limiting is disabled in the default compose config (--rate-limit=0) to avoid 429 responses during high-volume telemetry shipping.
• --profile full enables the telemetry agent and uses extra container capabilities for PID discovery. If it’s blocked in your environment, use --profile mock and run mcpdrill-agent on the MCP host instead.

Option B: From Source

Prerequisites: Go 1.24+, Node.js 18+ (for Web UI)

git clone https://github.com/bc-dunia/mcpdrill.git
cd mcpdrill
make dev

This starts all services on loopback (no auth required):

Stop with make dev-stop. View logs with make dev-logs.

For the Web UI in development mode (with hot reload):

cd web/log-explorer
npm install
npm run dev

Open http://localhost:5173. Or build and embed into the server: make frontend then restart the server to access at /ui/logs/.

Run a Test

Click "New Run" in the Web UI to create and start a test using the built-in wizard.

Or via CLI:

curl -X POST http://localhost:8080/runs \
  -H "Content-Type: application/json" \
  -d "{\"config\": $(cat examples/quick-start.json)}"

# Start the run (replace with your run_id from the response)
curl -X POST http://localhost:8080/runs/{run_id}/start

Results appear in real time in the Web UI dashboard.

Note: The quick-start config targets http://127.0.0.1:3000/mcp (IPv4). Using localhost may fail due to IPv6 resolution on some systems.

Optional: Server Resources (CPU/Memory)
The Web UI can display server-side CPU/memory metrics, but this requires the optional mcpdrill-agent. The quick-start config has server_telemetry.enabled: false by default. To enable it, see Server Telemetry Agent section below.

Key Features

Mock Server

MCP Drill includes a built-in mock MCP server with 27 tools for isolated testing without external dependencies.

Available Tools

Verify It Works

Make sure the mock server is running (e.g. docker compose --profile mock up -d or make dev) before calling localhost:3000.

# List available tools
curl -X POST http://localhost:3000/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"tools/list","id":1}'

# Call a tool
curl -X POST http://localhost:3000/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"tools/call","params":{"name":"fast_echo","arguments":{"message":"hello"}},"id":2}'

See examples/quick-start.json for a complete test configuration. Full tool documentation: Tool Testing Guide.

Server Telemetry Agent (Optional)

Note: MCP Drill works perfectly without the telemetry agent. This is an optional add-on for correlating client-side load test metrics with server-side resource usage.

Monitor server-side metrics (CPU, memory, threads, file descriptors) from your MCP server during load tests.

┌─────────────────────────────┐          ┌─────────────────────────┐
│   Your MCP Server Host      │          │   Control Plane         │
│                             │          │                         │
│  MCP Server   mcpdrill-agent│─────────▶│  Correlates metrics     │
│  (port 3000)       │        │  HTTPS   │  with test runs         │
│       ▲            │        │          │                         │
│       └── monitors─┘        │          │                         │
└─────────────────────────────┘          └─────────────────────────┘

Quick Setup

# Build the agent
make agent

# Enable on control plane
./mcpdrill-server --addr :8080 --enable-agent-ingest --agent-tokens "secret-token"

# Run agent on your MCP server host
./mcpdrill-agent \
  --control-plane-url http://localhost:8080 \
  --agent-token "secret-token" \
  --pair-key "my-test" \
  --listen-port 3000

Link with Test Runs

{
  "scenario_id": "capacity-test",
  "server_telemetry": {
    "enabled": true,
    "pair_key": "my-test"
  },
  "target": { "url": "http://localhost:3000/mcp" }
}

What is pair_key?
The pair key links metrics from a specific server with your test runs. Use the same value for --pair-key when starting the agent and in your run config's server_telemetry.pair_key to correlate server-side resource metrics (CPU, memory) with client-side load test results.

See Agent Telemetry Guide for full details.

API Quick Reference

# Create a run
curl -X POST http://localhost:8080/runs -H "Content-Type: application/json" -d @config.json

# Start / Stop / Emergency stop
curl -X POST http://localhost:8080/runs/{run_id}/start
curl -X POST http://localhost:8080/runs/{run_id}/stop -d '{"mode":"drain"}'
curl -X POST http://localhost:8080/runs/{run_id}/emergency-stop

# Status and events
curl http://localhost:8080/runs/{run_id}
curl -N http://localhost:8080/runs/{run_id}/events

Docker

Images

Published to GitHub Container Registry on every release:

# Pull a specific release (recommended)
docker pull ghcr.io/bc-dunia/mcpdrill/server:0.1.3

# Or use :latest for the most recent release
docker pull ghcr.io/bc-dunia/mcpdrill/server:latest

Note: Git tags use a v prefix (v0.1.3), but Docker image tags do not (0.1.3). Multi-arch images are published for linux/amd64 and linux/arm64.

Compose Profiles

docker compose up -d                       # server + worker
docker compose --profile mock up -d        # + mock server
docker compose --profile full up -d        # + mock server + agent
docker compose up -d --scale worker=5      # scale workers

Build Images

make docker-build                          # all images
make docker-build VERSION=v0.1.3           # with version tag
make docker-server                         # server only

Images are automatically built and published to GHCR on tag push via GitHub Actions.

Documentation

Examples

Pre-built configs in examples/:

• quick-start.json - Minimal test against mock server
• tool-testing/ - Tool validation scenarios
• scenarios/ - SSE and streaming tests
• multi-node/ - Docker Compose and Kubernetes

License

MIT License - see LICENSE for details.

MCP Drill was originally created to validate the production stability of Peta, an MCP control plane and runtime. If you're building MCP infrastructure that needs to scale, Peta handles the hard parts — routing, auth, observability — so you can focus on your agents.