Let an agent hunt down performance waste
Every codebase has optimizations nobody’s gotten around to. Auto-Tune is an Agent0 routine that reads your continuous profiling data every day, compares it to the day before, and tells you exactly what changed.
Trusted by
From profiling data to a mergeable PR
Auto-Tune sits on top of the profiling data that is already collected. Point it at a project, set a schedule, and it takes it from there.
Built on the agent stack your team already uses
Auto-Tune is built on the same protocols powering the rest of the AI-agent ecosystem.
GitHub
Reads code, PRs, commits, CI. Writes pull requests with the fix and the profiling evidence
Linear
Writes up findings that don’t warrant a PR yet as issues.
Any MCP server
Reads continuous profiling, trace, and metric data and acts on any MCP service
And more
OpenTelemetry-native. Connect any profiler or tool that speaks OTLP
Frequently asked questions
Auto-Tune is a feature that autonomously analyzes your continuous profiling data on a schedule, identifies performance and cost optimizations, and opens pull requests for the changes.
It compares your continuous profiling data (CPU, memory, and GPU), looking for architectural changes, regressions, and inefficiencies that are likely to produce a measurable improvement.
No. Auto-Tune opens a pull request with its proposed fix and the profiling evidence behind it. A human on your team still reviews and merges.
Any language or runtime that emits continuous profiling data as OpenTelemetry profiles, including Go, Rust, Java, Python, and Node.js.
Auto-Tune is built on the same agent-routine as Claude Code, and runs through an MCP with OAuth-based authentication.
Let the agent do the digging
Every codebase has optimizations nobody's gotten around to. Auto-Tune finds them, explains them in plain language, and opens the PR