Agents & skills

Agents are configuration, not code

Pick a model, write its instructions, attach knowledge, skills and tools — a change goes live without a deploy.

Configuration, not code

An agent's model, instructions and tools are settings, not a deployment.

Sandboxed skills

A skill's script runs in an isolated container, never inside the API process.

Tools over MCP

Agents call external tools and services through the Model Context Protocol.

Memory that cleans itself up

Agents remember users across conversations. In the background, kovac merges duplicate memories, corrects outdated facts and expires stale ones, and any memory can be inspected or deleted.

Plugins, versioned in git

Agents, skills and shared instructions ship as a plugin your team reviews in git, the same way it reviews code.