Fiji MCP Server

Ask your AI Agent in plain English to use Fiji / ImageJ to quickly analyze microscopy image also set pipelines from Cursor, Claude, Gemini, etc. You simply paste the image or ask it to navigate to the correct file.

The goal is that the AI agent should use the right ImageJ plugin, see what you are seeing and then verify its own results by writing codes without relying on vibes. I plan to setup SKILL and AI plugins in future. Would be happy to collaborate.

<!-- Demo images: absolute raw.githubusercontent.com URLs + Markdown tables (PyPI does not host ./demo_output; raw HTML <img> is less reliable in Warehouse). -->

---

See it in action

"Open the image, apply a Gaussian blur, show me before and after."

Before	After

"Threshold the bright spots, outline each object, report area and circularity."

Input	Outlined objects

#	Area	Circularity
1	1052	0.89
2	2840	0.72
3	641	0.91
4	1902	0.68

"Skeletonize the mask and summarize branches per tree."

Mask	Skeleton

Tree	Branches	Junctions
1	14	6
2	9	7

---

Get started in 3 steps

1 — Install

pip install fiji-mcp-server

You need Python 3.10+, Fiji installed on your machine, and Java (required by PyImageJ). See quickstart if anything needs clarification.

2 — Connect to your AI app

Replace /Applications/Fiji with your actual Fiji folder (the one containing jars/ and plugins/).

App	One command
Claude Desktop	fiji-mcp-install install claude-desktop --fiji-path /Applications/Fiji
Cursor	fiji-mcp-install install cursor --fiji-path /Applications/Fiji
Claude Code	fiji-mcp-install install claude-code --fiji-path /Applications/Fiji
Gemini CLI	fiji-mcp-install install gemini --fiji-path /Applications/Fiji
Windsurf	fiji-mcp-install install windsurf --fiji-path /Applications/Fiji

Then restart the app.

3 — Verify it works

In chat, type:

Run the Fiji MCP health_check tool

You should get back the Fiji version and mode. First startup takes 30–90 seconds while the JVM loads — that's normal.

---

What to ask

Once connected, just describe what you want:

"Open ./images/cells.tif and tell me the dimensions."

"Apply a Gaussian blur with sigma 4 and show me the result."

"Count the bright objects and give me their areas."

"Search for ImageJ commands related to 'threshold'."

"Open the image, subtract background, threshold, count particles — show me a screenshot after each step."

No macro knowledge needed. The assistant finds the right Fiji plugin, runs it, and can show you a screenshot to verify.

---

Available tools (19 total)

Category	Tools
Run & I/O	health_check run_macro run_batch_macros open_image save_image
Screenshots	screenshot_fiji — full screen, active image, or results table
Discover plugins	list_all_commands search_commands describe_plugin list_extensions
Image info	list_open_images get_image_info
Workflows	run_workflow — chain steps with screenshot verification
Results	parse_macro_output compare_screenshots list_macro_templates get_macro_template
Session	get_session_trace clear_session_trace

---

Documentation

Quick start	Install, configure, verify — step by step
All tools	What every tool does and when to use it
Configuration	Environment variables and troubleshooting
Architecture	How the pieces fit together

---

Author: Suraj Sahu · UC Merced Physics · ssahu2@ucmerced.edu

Related: cellpose_mcp · PyImageJ · FastMCP

License: BSD-3-Clause