AnyCap is a capability layer for AI agents — providing a clean, reusable way to generate images, analyze media, search the web, handle files, and produce usable outputs through one CLI.
It integrates with existing agent setups like Claude Code, Codex, and Cursor, without requiring replacement or lock-in. Instead of relying on fragmented APIs or SaaS tools, capabilities can be directly installed and executed inside the agent workflow.
This also opens up a more practical agent harness workflow — where agents operate with real capabilities, not just prompts.
Going forward, more use cases will be explored around real-world execution, multimodal tasks, and collaborative agent workflows.
GitHub: https://github.com/anycap-ai/anycap
Feedback from people building agents, tools, or infra is highly appreciated.