PixelleBLOCK
An Open-Source Multimodal AIGC Solution based on ComfyUI + MCP + LLM https://pixelle.ai
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
🎨 Pixelle MCP - Omnimodal Agent Framework
English | 中文
✨ An AIGC solution based on the MCP protocol, supporting both local ComfyUI and cloud ComfyUI (RunningHub) modes, seamlessly converting workflows into MCP tools with zero code.
https://github.com/user-attachments/assets/65422cef-96f9-44fe-a82b-6a124674c417
📋 Recent Updates
- ✅ 2025-09-29: Added RunningHub cloud ComfyUI support, enabling workflow execution without local GPU and ComfyUI environment
- ✅ 2025-09-03: Architecture refactoring from three services to unified application; added CLI tool support; published to PyPI
- ✅ 2025-08-12: Integrated the LiteLLM framework, adding multi-model support for Gemini, DeepSeek, Claude, Qwen, and more
🚀 Features
- ✅ 🔄 Full-modal Support: Supports TISV (Text, Image, Sound/Speech, Video) full-modal conversion and generation
- ✅ 🚀 Dual Execution Modes: Local ComfyUI self-hosted environment + RunningHub cloud ComfyUI service, users can flexibly choose based on their needs
- ✅ 🧩 ComfyUI Ecosystem: Built on ComfyUI, inheriting all capabilities from the open ComfyUI ecosystem
- ✅ 🔧 Zero-code Development: Defines and implements the Workflow-as-MCP Tool solution, enabling zero-code development and dynamic addition of new MCP Tools
- ✅ 🗄️ MCP Server: Based on the MCP protocol, supporting integration with any MCP client (including but not limited to Cursor, Claude Desktop, etc.)
- ✅ 🌐 Web Interface: Developed based on the Chainlit framework, inheriting Chainlit's UI controls and supporting integration with more MCP Servers
- ✅ 📦 One-click Deployment: Supports PyPI installation, CLI commands, Docker and other deployment methods, ready to us
124abd9690e5OBSERVED · 2026-09-25Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add pixelle -- uvx pixelle
{
"mcpServers": {
"pixelle": {
"command": "uvx",
"args": [
"pixelle"
]
}
}
}Exposed tools (4)
3 read · 0 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
get_workflow_tool_detail | read | |
list_workflows_tool | read | |
reload_workflows_tool | read | |
remove_workflow_tool | destructive |
Trust audit
BLOCKgrade D · trust 69/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (3 observation(s))
- Network
- declared (7 observation(s))
- Shell
- declared (4 observation(s))
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (8)
exec(func_def, {importlib.import_module(module_name_with_ext)
remove_workflow_tool
streamable-http
hash_object = hashlib.md5(content.encode())
f.write(base64.b64decode(base64_str))
docs/easy-workflow.png
docs/t2i_by_flux_turbo.png
Gates applied: no_behavioural_pass.
124abd9690e5full audit observations/trust-audit/mcp-server/aidc-ai__pixelle.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-09-25 | 124abd9690e5 | BLOCK | D | 69 | first audit |
Questions
What is the Pixelle MCP server?
An Open-Source Multimodal AIGC Solution based on ComfyUI + MCP + LLM https://pixelle.ai
What tools does Pixelle expose?
4 in total: 3 read-only, 0 that write, and 1 that can delete or overwrite (remove_workflow_tool). Every one is listed on this page with its risk.
Is Pixelle safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (69/100) and found 1 critical or high issue in the source. Each one is listed on this page with the file and line it is on. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does Pixelle need?
No credential environment variables were found in its source, so it appears to need none.
How does Pixelle run?
It speaks streamable-http, so it runs as a service you connect to over the network. It is published on PyPI as pixelle.
How current is this page?
The grade is for one exact copy of the source (124abd9690e5), read on 2026-09-25. The repository is watched and re-audited when it changes.