Yes. Claude Code can analyze images you provide, including screenshots, diagrams, and mockups. Anthropic’s Claude Code documentation describes image-analysis workflows for these kinds of visuals. This means Claude can work with an image you supply; it does not mean Claude Code automatically sees your screen or can inspect an image you have not made available to it. (code.claude.com)
How to provide an image
For a project image, give Claude Code the image’s file path and say what you want it to examine. Anthropic’s Claude Code guidance describes sharing a visual mock by providing its file path. (anthropic.com)
For example, place a screenshot in your project and ask:
Review
./screenshots/login.png. What looks misaligned, and which component or CSS should I inspect first?
Use a path to a file available to the session. The cited guidance does not specify platform or terminal conditions for this file-path method, so don’t assume that paste or drag-and-drop controls work the same way across Claude Code and other interfaces.
A supplied image is different from a live screen. Claude Code has a separate computer-use feature for interacting with a computer; it is not the same as asking about a screenshot file. (code.claude.com)
Claude Code vs. the production Claude API
These are separate workflows:
- Claude Code: an interactive coding tool. Provide the image to the session, for example by giving it a project image’s path.
- Claude API: your application or script sends an image in a Messages API request. Anthropic documents image content blocks with base64 data, an image URL, or a Files API
file_id. Its current Vision guide shows a Python SDK example usingclaude-opus-5-5; check the documentation for current model availability, limits, and costs before deploying. (platform.claude.com)
Here is a Python example using the documented base64 image-block structure and SDK call. Set ANTHROPIC_API_KEY in your environment and install the SDK with pip install anthropic. This is a direct API request—not a Claude Code CLI command. (platform.claude.com)
import base64
from pathlib import Path
import anthropic
image_data = base64.standard_b64encode(
Path("./screenshots/login.png").read_bytes()
).decode("utf-8")
client = anthropic.Anthropic() # reads ANTHROPIC_API_KEY
message = client.messages.create(
model="claude-opus-5-5",
max_tokens=1024,
messages=[
{
"role": "user",
"content": [
{
"type": "image",
"source": {
"type": "base64",
"media_type": "image/png",
"data": image_data,
},
},
{"type": "text", "text": "Describe the visible login layout."},
],
}
],
)
for block in message.content:
if block.type == "text":
print(block.text)
The API’s image-input format should not be mistaken for Claude Code’s image-sharing controls. Use Anthropic’s current Vision documentation for API details that may change, including supported models and request limits. (platform.claude.com)
What image analysis can help with
Anthropic’s Claude Code documentation identifies screenshots, diagrams, and mockups as image-analysis use cases. You might ask Claude to describe a UI screenshot, explain a flowchart, or relate a visible issue to code in your project. Treat the response as analysis to review, not a guarantee that every visual detail was interpreted correctly. (code.claude.com)
Bottom line: Claude Code can analyze images you supply. For a project task, give it the image’s accessible file path and a focused question. If you are building an application that sends images to Claude, use the separate Messages API workflow.
Sources
Common workflows — Claude Code Docs
Claude Code Best Practices — Anthropic