cap-evolve
by Community
Optimize any AI agent’s skills, tools/MCP, and prompts against your own evals.
OSS
cap-evolve
Added 1 Oct 2026
Overview
cap-evolve is a Python-based observability tool that optimizes an AI agent's skills, tools/MCP, and prompts against user-defined evals. It uses evaluation results to guide adjustments to agent components.
Best for
Best for
Developers building custom AI agents who want to optimize components against their own evaluation suites.
Use cases
- Tune agent prompts against custom evals
- Optimize tool and MCP selections for specific tasks
- Iterate on agent skills using evaluation feedback
Notes
cap-evolve is a Python-based observability tool that optimizes an AI agent’s skills, tools/MCP, and prompts against user-defined evals. It uses evaluation results to guide adjustments to agent components.
65 stars on GitHub. Last updated 2026-10-01. Licensed Apache-2.0.
Use cases
- Tune agent prompts against custom evals
- Optimize tool and MCP selections for specific tasks
- Iterate on agent skills using evaluation feedback
Pros
- Uses your own evals, so optimization aligns with your criteria
- Covers skills, tools/MCP, and prompts in one tool
- Open source and community maintained
Cons
- Small community with 65 stars suggests limited adoption
- Requires existing evals to function
- Python only, which may not fit non-Python stacks
Indexed from awesome-llmops and enriched against its public facts.
Pros
- Uses your own evals, so optimization aligns with your criteria
- Covers skills, tools/MCP, and prompts in one tool
- Open source and community maintained
Cons
- Small community with 65 stars suggests limited adoption
- Requires existing evals to function
- Python only, which may not fit non-Python stacks
Get the free Developer’s Field Guide
A 27-page field guide to the AI coding workflow with Claude. Claude Code, MCP servers, the prompt patterns that work, and what to delegate. Free.
Enter your work email. We send it straight over, plus a few short notes worth knowing. Unsubscribe any time.