· ai-devtools
HarnessAgent supports image inputs across all harness adapters like Claude Code, Codex, and GitHub Copilot
Developers can now provide images to the AI SDK HarnessAgent layer through every harness adapter.
This capability allows users to send visual data alongside textual prompts using the established AI SDK message format. Specifically, this involves a content array that combines parts designated as `{ type: 'text' }` and `{ type: 'file' }`.
The harnesses utilize their integrated vision models to process these incoming image inputs.
Support for this visual input is comprehensive, spanning all currently available adapters.
These adapters include Claude Code, Cline, Codex, Cursor, Deep Agents, fx, GitHub Copilot, Grok Build, OpenCode, and Pi.
To begin leveraging this functionality, developers should consult the HarnessAgent image inputs documentation.
