mandelbulber2-mcp
MCP server for inspecting, cloning, patching, validating, and rendering preview images from Mandelbulber 2.34 fractal scene files, with safety limits and verified execution.
README
Mandelbulber 2 MCP
A fail-closed MCP server for inspecting, cloning, safely patching, validating, and rendering preview images from Mandelbulber 2.34
.fractscenes.
This first release deliberately controls a small, verified part of Mandelbulber. It works from trusted templates, preserves scene bytes outside approved edits, invokes the CLI without a shell, and verifies the generated PNG instead of trusting Mandelbulber's exit code or Image saved message.
What it can do
| Tool | Purpose |
|---|---|
mandelbulber_probe |
Verify the configured executable, SHA-256, and exact 2.34 version output |
fract_inspect |
Read lexical header, sections, records, duplicates, scripts, and scene hash |
fract_clone |
Create a byte-identical managed copy with a provenance manifest |
fract_patch_safe |
Atomically edit existing detail_level, image_width, or image_height records |
fract_validate |
Verify trusted-template provenance and the conservative static preview subset |
fract_render_preview |
Render one bounded CPU/PNG still and verify signature, IHDR, dimensions, and hash |
Animations, queues, netrender, GPU flags, runtime scripts, custom formula code, arbitrary parameter overrides, arbitrary scene paths, and final-resolution renders are not exposed.
Requirements
- Node.js
24.12.x(the repository includes.nvmrcfor nvm-windows) - Mandelbulber
2.34 - A read-only trusted template directory
- A separate writable work directory
The executable is administrator-configured; MCP clients cannot supply a command, path, argv, environment variable, or output destination.
Quick start
nvm use 24.12.0
npm ci
npm run check
npm run build
Create two disjoint directories, copy approved .fract templates into the template root, and configure the server:
$env:MANDELBULBER_EXECUTABLE = "C:\Program Files\Mandelbulber2\mandelbulber2.exe"
$env:MANDELBULBER_EXECUTABLE_ROOT = "C:\Program Files\Mandelbulber2"
$env:MANDELBULBER_EXECUTABLE_SHA256 = "<lowercase sha256>"
$env:MANDELBULBER_TEMPLATE_ROOT = "C:\mandelbulber-mcp\templates"
$env:MANDELBULBER_WORK_ROOT = "C:\mandelbulber-mcp\work"
node .\dist\src\index.js
The process uses stdio for MCP traffic. Keep stdout reserved for the protocol.
MCP client configuration
Build first, then add a stdio server entry to your MCP client. Use absolute paths:
{
"mcpServers": {
"mandelbulber2": {
"command": "C:\\Program Files\\nodejs\\node.exe",
"args": ["C:\\path\\to\\mandelbulber2-mcp\\dist\\src\\index.js"],
"env": {
"MANDELBULBER_EXECUTABLE": "C:\\Program Files\\Mandelbulber2\\mandelbulber2.exe",
"MANDELBULBER_EXECUTABLE_ROOT": "C:\\Program Files\\Mandelbulber2",
"MANDELBULBER_EXECUTABLE_SHA256": "<lowercase sha256>",
"MANDELBULBER_TEMPLATE_ROOT": "C:\\mandelbulber-mcp\\templates",
"MANDELBULBER_WORK_ROOT": "C:\\mandelbulber-mcp\\work"
}
}
}
}
Logical template paths start with templates/; managed scenes are returned as workspaces/<uuid>/<name>.fract. A normal client workflow is:
probe → inspect template → clone → patch with expected SHA-256 → validate → render preview
Safety limits
- Exact supported CLI version:
2.34 - Preview dimensions:
5–512per side, at most262,144pixels - Default preview:
320×240; format: PNG only - One render at a time; fixed 120-second timeout
- Maximum scene size: 8 MiB; bounded stdout/stderr capture: 64 KiB each
- Managed clones only; expected-hash concurrency checks and replayable edit manifests
- No shell invocation; timeout/cancellation terminates the process tree
Mandelbulber 2.34 on Windows can still resolve known profile directories outside environment-variable overrides. For production isolation, run this server and Mandelbulber under a dedicated restricted Windows account and grant that account access only to the executable, trusted templates, and managed work root.
Project status
0.1.0 is a complete, deliberately narrow MVP. The automated suite covers contracts, filesystem boundaries, parser behavior, provenance replay, process timeout/capture, PNG verification, SDK tool registration, and the multi-tool MCP workflow. A gated integration test also exercises an installed Mandelbulber 2.34 executable.
The source investigation and architecture rationale live in docs/. Start with implementation-plan-2.34.md, 2.34-baseline.md, and security-model.md.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.