Aspen-MCP (EPEL-SNU)
MCP serverCommunity
Research server with 42 typed Aspen Plus tools, a spec-to-flowsheet builder, EO optimisation and costing. By EPEL-SNU.
6 starsUpdated 4 weeks agoGitHub data from 5 Nov 2026
At a glance
- Type
- MCP server
- Works with
- Aspen Plus
- How it connects
- Drives a running copy of the tool
- Access
- Can modify models
- Tool versions
- Developed on Aspen Plus V14
- Operating systems
- Windows
- Needs a Aspen Plus licence
- Yes
- Transport
- Local (stdio), Remote (HTTP)
- AI clients documented
- Claude Desktop, OpenClaw
- Licence
- MIT
- Language
- Python
What it can do
- Gives the model typed tools for reading stream data, setting block conditions and running cases, so it doesn't have to guess COM variable paths.
- Builds a runnable .bkp from a declarative flowsheet spec, starting from a blank template.
- Bridges to equation-oriented optimisation and a techno-economic analysis that estimates CAPEX, OPEX and annualised profit for a converged flowsheet.
- Ships the benchmark harness and result logs from the authors' study comparing tool calls with raw code generation.
Limitations
- Windows only and needs a licensed Aspen Plus; the authors describe it as research code.
- EO optimisation needs an EO-prepared model, and the TEA tool needs a converged one.
- The evaluation and demo scripts expect an OpenAI-compatible LLM endpoint, such as a local vLLM server.
Install
pip install -r requirements.txtpython src/aspen_mcp_server.py --transport http --host 127.0.0.1 --port 8765MCP servers and skills run code on your computer. Review the source before installing.
AI-generated models, drawings and results must be checked by a qualified engineer before use.