[BUG] - Read tool: Claude cannot interpret or see image content when reading image files
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
When using the Read tool to read image files, Claude is unable to interpret or see the image content. The tool executes successfully, but Claude cannot analyze, describe, or respond to prompts about what is in the image.
In contrast, when the same image is pasted directly via Ctrl+V/Cmd+V, Claude can see and interpret the image content normally.
Result: Claude can see and interpret the image, describing its contents accurately.
What Should Happen?
According to the Read tool description:
"This tool allows Claude Code to read images (eg PNG, JPG, etc). When reading an image file the contents are presented visually as Claude Code is a multimodal LLM."
Claude should be able to interpret the image content and respond to prompts about what is in the image — the same way it can when images are pasted directly.
Error Messages/Logs
None
Steps to Reproduce
- Have a valid PNG/JPG image file on disk (e.g., /tmp/test_image.png)
- Verify the file is valid:
file /tmp/test_image.png
# Output: PNG image data, 2478 x 832, 8-bit/color RGBA, non-interlaced
- Ask Claude Code to read and describe the image:
- "Read /tmp/test_image.png and tell me what's in it"
- Or reference via @/tmp/test_image.png
- Claude uses the Read tool, which returns successfully
- Ask Claude what it sees in the image
Result: Claude cannot see or interpret the image content. It cannot describe what's in the image or answer questions about it.
- Now paste the same image via Cmd+V/Ctrl+V
- Ask Claude what it sees
Claude Model
Opus
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
2.1.9
Platform
AWS Bedrock
Operating System
macOS
Terminal/Shell
iTerm2
Additional Information
_No response_
This issue has 7 comments on GitHub. Read the full discussion on GitHub ↗