Evo
Turns work on a codebase into an autoresearch loop: the agent decides what to measure, sets up a benchmark and runs tree search with parallel sub-agents
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Runs experiments that execute code
- Can operate in remote sandboxes via third-party providers
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/evoDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Run one by one in the Claude Code chat
/plugin marketplace add evo-hq/evo
/plugin install evo@evo-hq-evoInstall the CLI: uv tool install evo-hq-cli, then wire it to a host: evo install claude-code (or codex, cursor, opencode).
Other ways from the author
uv tool install evo-hq-cliFor remote backends install with an extra, for example 'evo-hq-cli[modal]'.
This is third-party code. Review the repository files before installing.
What it does
Evo helps an agent improve code through a measurable loop. First it decides what to measure and instruments the project with a benchmark, then it runs a search over a tree of solutions with several sub-agents and compares results. Experiments can run locally or in remote sandboxes from various providers. It installs as a CLI that adds a plugin to the chosen agent, and the invocation depends on the host.
Who it is for. For developers who want to improve code by measurable metrics rather than by feel.
Good fit when
- You need to optimize code against a clear metric
- You want to automatically explore many solution variants
- You need a repeatable benchmark for comparison
Not a fit when
- The task cannot be expressed as a measurable metric
- You cannot run many experiments that execute code
Example request
With Evo set up a benchmark on this function's speed and explore optimization variantsLimitations
The loop runs experiments that execute code, which needs resources, and remote sandboxes rely on third-party providers and their keys. The README describes installing extras for each backend.
How to disable. Remove the evo-hq-cli tool via your package manager and uninstall the evo plugin from your agent.
Security check
- Runs experiments that execute code
- Can operate in remote sandboxes via third-party providers
README in short
The README describes Evo as an autoresearch loop over code: choosing a metric, a benchmark and tree search with parallel sub-agents. It installs via uv tool install evo-hq-cli and an evo install command for the chosen host, with host-specific invocation. Local and remote backends are supported through extras. It covers plugin updates and running a dashboard.
FAQ
Where do experiments run?
Locally or in remote sandboxes: Modal, E2B, Daytona, AWS, Azure or over SSH. Each backend installs its own extra.
How is it invoked?
The syntax depends on the host, for example /evo: in Claude Code.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything