performance_engineer — Claude Code subagent: analysis and setup
The performance_engineer Claude Code subagent finds application, API, database, and infrastructure bottlenecks from measurements. It builds an optimization plan with baselines, experiments, and regression checks.
no license ★ 776 updated 2026-11-12
.claude/agents/performance_engineer.mdWhat performance_engineer does
The subagent reviews performance evidence and returns a Markdown report. It records the baseline, workload model, and evidence quality, then ranks bottleneck hypotheses and experiments with their expected trade-offs. The report covers CPU, memory, I/O, networking, concurrency, and databases.
It also sets out a benchmark protocol, acceptance thresholds, regression tests, and rollback steps. Measured facts stay separate from hypotheses and recommendations; benchmark results and production traffic are never fabricated. Benchmark output, profiles, traces, resource metrics, and architecture context can be supplied as supporting evidence, but they are optional.
What it helps with
- locate API bottlenecks from measurements
- rank experiments for a latency target
- assess optimization trade-offs across CPU and memory
- define acceptance thresholds and regression checks
Example prompts for the agent
- Analyze these API performance measurements and create an optimization plan with baselines and regression checks.
- Write a database performance report from these profiles and traces.
- Compare the measured workload with the latency budget and rank the next experiments.
How to install performance_engineer
Manually: put the files in place
.claude/agents/performance_engineer.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/performance_engineer.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/performance_engineer.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about performance_engineer
How do I install performance_engineer in Claude Code?
Add the subagent file at .claude/agents/performance_engineer.md.
What does performance_engineer do?
It examines performance evidence for an application, API, database, or infrastructure and produces a Markdown report with bottleneck hypotheses, ranked experiments, and regression controls.
What evidence can performance_engineer use?
Provide a workload target, latency or throughput budget, environment, and constraints. Benchmark output, profiles, traces, resource metrics, and architecture context are optional supporting evidence.