[BUG] mcp__claude-in-chrome__computer stores temp file path in base64 image data field instead of actual image data
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?
Bug Report
Tool: mcp__claude-in-chrome__computer (screenshot tool)
### What happens
When the computer tool captures a screenshot, the session .jsonl file records the result with a temp file path (e.g.
/var/folders/.../T/steer/c9060e38png) in the source.data field instead of actual base64-encoded image data.
### Effect
When Claude Code resumes that session in a later run, the Anthropic API immediately rejects the request with:
400 {"type":"error","error":{"type":"invalid_request_error","message":"Could not process image"}}
The session is permanently broken because:
- The
source.typeisbase64but thedatafield contains a file path string - The temp file is deleted between runs, so the data can't be recovered
### Workaround
Manually edit the .jsonl session file and replace the bad image entry:
```json
{"type": "image", "source": {"type": "base64", "media_type": "image/png", "data": "/var/folders/..."}}
with a text placeholder:
{"type": "text", "text": "[screenshot unavailable — temp file was deleted]"}
Environment
- Claude Code version: 2.1.77
- Platform: macOS (darwin arm64)
- MCP server: claude-in-chrome (in-process)
What Should Happen?
The computer tool should embed the actual base64-encoded PNG data in the data field, not a path to a temp file.
Error Messages/Logs
400 {"type":"error","error":{"type":"invalid_request_error","message":"Could not process image"}}
Steps to Reproduce
capture a screenshot, end session, try to continue session
Claude Model
Sonnet (default)
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.1.77 (Claude Code)
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
iTerm2
Additional Information
_No response_
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗