uniprof — Claude Code MCP server: what it does, how to install
Profiles CPU usage across runtimes without code modifications or extra packages. Gives AI assistants structured data on execution bottlenecks.
claude mcp add uniprof -- npx -y uniprofWhat uniprof does
The tool wraps specialized profilers for runtimes like Python 3.7+ (py-spy), Node.js 14+ (0x), and Java 8+ (async-profiler), alongside Ruby, PHP, and compiled binaries. It selects the target profiler based on command arguments, aggregates varied raw outputs into a unified model, and calculates hotspots.
Isolation matters. uniprof defaults to isolated Docker environments to sidestep complex permission setups on local machines. Developers can pass --mode host if the profilers exist locally. Container execution fails on macOS Mach-O binaries, because Apple Instruments requires the native host.
Compiled applications need DWARF symbols and frame pointers to resolve function names. Windows host environments are excluded, though WSL2 operates without native binary profiling.
What it helps with
- profile CPU consumption without editing project files or dependencies
- identify performance hotspots across Python, Node.js, and JVM runtimes
- normalize varied profiler outputs into a single structured report
- record execution traces for flamegraph inspection
Example prompts for the agent
- Profile the execution of
python app.pyand summarize the slowest methods - Record a CPU profile for
node server.jsand write it toprofile.json - Inspect
profile.jsonto pinpoint where the process spends execution time
How to install uniprof
Platform CLI
claude mcp add uniprof -- npx -y uniprofPlatform CLI
codex mcp add uniprof -- npx -y uniprofConfig file
{
"mcpServers": {
"uniprof": {
"command": "npx",
"args": [
"-y",
"uniprof"
]
}
}
}Config file
{
"servers": {
"uniprof": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"uniprof"
]
}
}
}Platform CLI
gemini mcp add uniprof npx -y uniprofStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about uniprof
How to install uniprof in Claude Code?
Run claude mcp add uniprof -- npx -y uniprof. If the CLI package is installed globally, you can also execute uniprof mcp install claudecode.
What does uniprof do?
It allows agents to inspect CPU usage across multiple programming languages without modifying source code. The server runs the appropriate profiling binary and extracts actionable execution hotspots.
Is Docker required to run uniprof?
By default, Docker is preferred for running language profilers inside isolated containers. You can bypass containers by passing --mode host when your host machine already contains the necessary profiling utilities.
