autochrome
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@autochromeApply Portra 400 film emulation to this photo and annotate it with a badge"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Autochrome
Autonomous photographic color science engine, MCP server, and real-time live preview studio for AI Agents (Claude Code, Cursor, Gemini CLI, Antigravity).

AI Agent Quick Reference & Protocol
If you are an AI Agent (Claude, Cursor, Gemini, Antigravity, OpenCode, Codex), Autochrome provides deterministic, professional-grade computer vision tools and color science heuristics to edit photos and annotate screenshots.
Standard 4-Step Agent Execution Lifecycle
When given a photo editing or screenshot annotation task, follow this standard sequence:
┌────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ CANONICAL AGENT EXECUTION LIFECYCLE │
├────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ 1. OPEN & PREVIEW ▶ open_image(image_path) │
│ • Initializes session and boots Native macOS Live Darkroom Window for user. │
│ │
│ 2. DIAGNOSE SCENE ▶ inspect_image() │
│ • Inspects Zone System exposure, lighting scenario, color cast, and depth. │
│ │
│ 3. COLOR GRADE & CV ▶ tune_image(), hsl_color_mixer(), color_wheels_grade(), portrait_retouch() │
│ • Dispatches deterministic CV tools (streamed to user's screen in real time). │
│ │
│ 4. FINALIZE & EXPORT ▶ export_image(output_path, format="PNG") │
│ • Saves the final developed master asset to disk. │
└────────────────────────────────────────────────────────────────────────────────────────────────────────┘Agent Rules of Engagement & Guardrails
Always Open Preview First: Calling
open_image(image_path)automatically launches the Native macOS Live Darkroom Preview window (or Web Studio) on the user's screen. Inform the user they can pressSpaceto hold-to-compare orSto toggle the split curtain slider.Never Guess Parameters Without Diagnosis: Always run
inspect_image()to checklighting_scenario(e.g.overcast_flat,golden_hour,harsh_sunlight),optical_depth_evaluation, andzone_system_analysis.Respect Optical Depth (Zero Fake Blur): If
has_natural_depth_of_fieldisTrue, DO NOT apply fake lens blur. Use tonal separation and micro-contrast instead to avoid edge halos.Preserve Skin Tones on the 120° Vectorscope Line: Always use
hsl_color_mixer(luminance_shifts={"orange": 12.0})to enhance skin radiance without altering backgrounds or clothing colors.One-Shot Master Development: For general enhancement without micromanagement, call
master_develop()to run the autonomous 5-phase critique and development loop.
Related MCP server: YOLO-FFMPEG-MCP
The Two Core Agent Superpowers
AI Agents can write code, browse the web, and run terminal commands. Autochrome gives them the missing visual superpowers:
┌─────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ AUTOCHROME AGENT SUPERPOWERS │
├────────────────────────────────────────────────────┬────────────────────────────────────────────────────┤
│ SUPERPOWER 01: AGENTS EDITING PHOTOS │ SUPERPOWER 02: AGENTS ANNOTATING SCREENSHOTS │
│ • Ansel Adams 11-Zone Luminance Calibration │ • CleanShot-Style macOS Dark Window Bezels │
│ • True Frequency Separation (100% Pore Retention) │ • Multi-Stop Slate Gradient Mesh Backdrops │
│ • Kodak Portra 400 & Analog Film Emulation │ • Curved Anti-Aliased Vector Arrows │
│ • 8-Channel HSL Mixing & 3-Way Color Wheels │ • Sleek Embedded Pill Badges & Label Chips │
│ • 100% Deterministic CV — Zero Diffusion Artifacts │ • 2.2x Floating Optical Detail Loupes │
└────────────────────────────────────────────────────┴────────────────────────────────────────────────────┘Live Darkroom Preview: Mac Native & Web Studio
Autochrome provides two preview experiences for monitoring AI agent edits in real-time, preferring native Mac preview by default:
┌─────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ DUAL LIVE PREVIEW MODES │
├────────────────────────────────────────────────────┬────────────────────────────────────────────────────┤
│ PREFERRED: NATIVE MAC PREVIEW WINDOW │ WEB BROWSER STUDIO (Cross-Platform) │
│ • Default mode on macOS │ • Access via browser at http://localhost:8000 │
│ • Dedicated floating darkroom window │ • Real-time WebSocket canvas streaming │
│ • Zero browser tabs or URL bar clutter │ • Interactive Split-Screen Curtain Slider (Key: S) │
│ • Native macOS Cocoa rendering & instant launch │ • Hold-to-Compare Original View (Key: Space) │
│ │ • Live Action Ledger & Quality Index telemetry │
└────────────────────────────────────────────────────┴────────────────────────────────────────────────────┘1. Launch Native Mac Preview (Default Preferred)
autochrome preview path/to/your/photo.jpgLaunches an isolated native macOS floating darkroom window that updates in real-time as the agent dispatches MCP tools.
2. Launch Web Browser Studio
autochrome preview path/to/your/photo.jpg --browserOpens the interactive Web Studio at http://localhost:8000 with interactive curtain slider and comparison controls.
Quick Start
1. Installation
# Clone repository
git clone https://github.com/hgayan7/autochrome.git
cd autochrome
# Create virtual environment and install
python3 -m venv .venv
source .venv/bin/activate
pip install -e .2. Run the Interactive Tour Demo
Watch the agent autonomously develop a photo and beautify a presentation screenshot with real-time live preview:
autochrome demoStreams the live transformations step-by-step with hold-to-compare and split curtain slider controls.
Connecting to AI Agents (Claude Code, Cursor, Gemini, Antigravity)
Autochrome provides both MCP Tools (the execution engine) and an Agent Skill (SKILL.md) (the artistic heuristics and photographic color science brain).
1. MCP Server Configuration (Tools)
Claude Desktop Configuration
Add Autochrome to your claude_desktop_config.json:
{
"mcpServers": {
"autochrome": {
"command": "/path/to/autochrome/.venv/bin/autochrome",
"args": ["mcp"]
}
}
}Cursor & Antigravity MCP Setup
In .cursor/mcp.json or Antigravity MCP settings:
{
"mcpServers": {
"autochrome": {
"command": "autochrome",
"args": ["mcp"]
}
}
}2. Agent Skill Setup (SKILL.md)
While MCP provides raw tool endpoints, the Autochrome Skill (SKILL.md) teaches AI agents how to edit like professional colorists and master photographers rather than blindly executing random adjustments.
Installing the Skill for AI Agents
# Antigravity & Gemini CLI
mkdir -p ~/.gemini/config/skills/autochrome
cp SKILL.md ~/.gemini/config/skills/autochrome/SKILL.md
# Claude Code
mkdir -p ~/.claude/skills/autochrome
cp SKILL.md ~/.claude/skills/autochrome/SKILL.md
# Cursor
mkdir -p .cursor/skills/autochrome
cp SKILL.md .cursor/skills/autochrome/SKILL.mdPython SDK Reference (For Code-Generating Agents)
Agents writing direct Python automation scripts can import Autochrome engine modules directly:
from autochrome.core.canvas import Canvas
from autochrome.mcp import tools
# 1. Load photo into session
tools.tool_open_image("input.jpg", launch_preview=True)
# 2. Diagnose scene lighting and metrics
metrics = tools.tool_inspect_image()
# 3. Apply color grading & retouching
tools.tool_tune_image(brightness=5, contrast=8, ambiance=20, highlights=-12, shadows=10, warmth=6)
tools.tool_hsl_color_mixer(luminance_shifts={"orange": 14.0})
tools.tool_portrait_retouch(face_spotlight=15, skin_smoothing=18, eye_clarity=30)
tools.tool_adjust_details(structure=12, sharpening=22)
# 4. Save result
tools.tool_export_image("output.png")Complete MCP Tool Catalog
Live Darkroom Preview & Session Control
Tool Name | Parameters | Description |
|
| Loads photo into active canvas and auto-launches Live Preview (Native macOS floating window or browser) |
|
| Explicitly launches or connects the real-time Live Darkroom Preview window |
|
| Exports final edited composite image to disk ( |
| none | Undoes the last editing action on the active canvas |
| none | Redoes the last undone action |
Photographic Color Science & Analog Film Stocks
Tool Name | Parameters | Description |
| none | Scene semantics, content genre (portrait, landscape, street, architecture), and adaptive recommendations |
|
| Autonomous scene-aware master development: diagnoses scene, adapts film stock, retouches portraits, and refines details |
|
| 13 authentic analog film stocks ( |
| none | Lists all 13 authentic film stock emulations with metadata and best use cases |
|
| CineStill-style crimson-orange specular halation bloom on intense light sources |
|
| Classic Michael Orton dreamy glow diffusion with micro-contrast retention |
|
| Chemical silver retention for a gritty, high-contrast, desaturated cinematic aesthetic |
|
| Dark Channel Prior (DCP) atmospheric haze and fog removal |
|
| Physical 2000K-12000K Planckian Blackbody white balance and green/magenta tint |
|
| Authentic density-dependent silver halide film grain peaking in Zone V midtones |
| none | Autonomous 5-phase master development loop elevating quality scores to 90+ Master Studio Quality |
| none | Computes 0-100 Photographic Aesthetic Index across Dynamic Range, Skin Radiance, Sharpness |
|
| Applies DaVinci Resolve / ARRI 3D LUT (.cube format) using trilinear interpolation |
|
| Exports active canvas grade as a standard .cube 3D LUT file |
|
| Eliminates purple & green lateral chromatic aberration color fringing |
|
| Compensates for optical lens light falloff (vignetting) |
|
| Merges bracketed exposures into high-dynamic-range photo using OpenCV Mertens fusion |
|
| Quad-corner document edge detection & flatbed perspective rectification |
|
| Carbon/Ray.so syntax-highlighted code snippet presentation mockup |
|
| Matches canvas color grade to a hero reference photo via CIELAB transfer |
|
| 16-Filter Library presets ('polaroid_70s', 'cinematic_pop', 'cross_process', etc.) |
| none | Lists all 16 available photo filters with descriptions |
|
| Creates official biometric ISO 35x45mm and US 2x2 studio passport photos (~75% face ratio) |
|
| Deterministic chroma-key background replacement with 100% pure foreground protection |
|
| Professional Frequency Separation: smooths skin tones while retaining 100% genuine pore texture |
|
| 4-Point perspective homography correcting converging lines |
|
| Camera calibration radial lens distortion removal (barrel / pincushion) |
|
| Ansel Adams 11-Zone photographic luminance calibration |
|
| Analyzes image histogram & lighting to compute and apply a custom tailored grade |
|
| 8-Channel targeted color grading (Orange for skin, Green for eyes, etc.) |
|
| 3-Way Lift/Gamma/Gain color wheels grading |
|
| Primary tone curve and local dynamic balancing (-100 to +100) |
|
| Fine micro-contrast texture pop and high-pass sharpening |
|
| Spline tone curves ( |
|
| Facial illumination, skin texture preservation, and iris pop |
|
| Content-aware inpaint patch healing brush |
|
| Localized radial control points |
Framing & Screenshot Studio Tools
Tool Name | Parameters | Description |
|
| Smart composition crop centered on face or focus point |
|
| macOS window bezel with traffic lights + mesh gradient backdrops |
|
| Anti-aliased curved vector arrow with embedded descriptive text pill badge |
|
| Anti-aliased curved/straight vector arrow |
|
| Spotlight rounded callout box with optional badge label |
|
| Semi-transparent marker highlight |
|
| Circular numbered tutorial step badge (1, 2, 3) |
|
| Zoom-in glass loupe showcasing fine detail |
|
| Gaussian privacy blur for passwords / sensitive keys |
|
| Mosaic pixelation for emails / names |
|
| Solid opaque blackout bar |
Vision & Inspection
Tool Name | Parameters | Description |
| none | Returns luminance, contrast, dynamic range, dominant color palette, and face bounding coordinates |
|
| Overlays a calibrated coordinate grid ( |
Testing
Run the automated test suite:
pytest -vLicense
MIT License • Created by @hgayan7
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Connectors
MCP server for Wan AI video generation
MCP server for building and testing AI agents with multi-model experimentation and insights.
MCP server connecting AI agents to non-custodial staking data across 130+ networks.
MCP server for Qwen Image 3 AI image generation
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceMCP server for visual feedback, video direction, and QA assertions on web pages, enabling AI agents to read, reply, and resolve annotations in real time.4MIT
- FlicenseNot gradedqualityDmaintenanceAn AI-powered MCP server for intelligent video processing, enabling automated analysis, transition effects, and quality assurance with multi-agent orchestration.2
- AlicenseAqualityBmaintenanceAn MCP server that lets AI agents visually perceive and edit video inside DaVinci Resolve Studio by rendering frames and manipulating the timeline.2MIT
- AlicenseNot gradedqualityBmaintenanceMCP server for OmniCut that provides 29 deterministic video editing tools (trim, transition, render, etc.) for AI agents to drive the video editor end-to-end.91ISC
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/hgayan7/autochrome'
If you have feedback or need assistance with the MCP directory API, please join our Discord server