AI-native engineering workspace for calculations, notebooks and report generation.
Agents can inspect a project, edit notebooks, run calculations and deliver DOCX/PDF reports.
Inspyro is an open source workspace for engineering teams that want local execution, notebook-first authoring, traceable report generation, and agent workflows in the same environment. The notebook, editor, explorer, and document pipeline are still there, but the product is now framed around outcomes: understand the project, run engineering calculations, and ship report artifacts.
- Agent-first engineering workflow: local agents can inspect project files, drive notebook execution, and update reports through the built-in MCP layer.
- Notebook runtime with real Python: calculations run on a real Jupyter kernel instead of a simulated playground.
- Report artifacts built into the workflow: notebooks can generate DOCX and PDF deliverables with the custom DOCX builder and Office math pipeline.
- Traceability by design: project files, notebook cells, figures, tables, and final artifacts stay connected.
- Local-first architecture: backend, frontend, desktop shell, and MCP server run on your machine.
Inspyro is designed around three promises:
- Understand the project: inspect the workspace, the notebook, the engineering inputs, and the template.
- Run engineering calculations: execute real Python, review variables, graphs, units, and dependency context.
- Ship report artifacts: generate DOCX/PDF outputs that remain connected to the source notebook and project files.
The repository includes a canonical example workspace in examples/structural-report-demo.
Demo flow:
- Open the example workspace in Inspyro.
- Review
inputs/beam_case.json,beam_design.py, andbeam_report.ipynb. - Run all notebook cells.
- Inspect the generated document in the
Documentview. - Connect an MCP client and repeat the same flow through the agent-facing surface.
Windows:
python -m venv venv_inspyro
.\venv_inspyro\Scripts\activate
pip install -r backend/requirements.txt
cd frontend
npm install
cd ..
.\restart_inspyro.ps1Linux / WSL:
python3 -m venv venv_inspyro
source venv_inspyro/bin/activate
pip install -r backend/requirements.txt
cd frontend && npm install && cd ..
./restart_inspyro.shOnce the app is running:
- Click
Start from exampleon the launcher, or open examples/structural-report-demo as your workspace. - Open
beam_report.ipynb. - Run the notebook.
- Open the generated DOCX/PDF artifact from the
Documentview.
Start the backend first, then run the local MCP server:
cd backend
python -m mcp_serverUseful variants:
python -m mcp_server --stdio
python -m mcp_server --json-response --stateless-http
python -m mcp_server --wait-for-backend 20Default MCP endpoint:
http://127.0.0.1:8100/mcp
Recommended agent flow:
- Read
inspyro://manifest. - Read
inspyro://guides/start-here. - Open or create a notebook-first session.
- Load the structural report demo workspace.
- Run calculations and export report artifacts.
flowchart LR
A["Human UI"] --> B["React Workspace Shell"]
G["External Agent"] --> H["Local MCP Server"]
B --> C["FastAPI Backend"]
H --> C
C --> D["Jupyter Kernel"]
C --> E["DOCX/PDF Pipeline"]
C --> F["Workspace Files"]
E --> I["DOCX / PDF Artifacts"]
- CONTRIBUTING.md: contribution workflow and checks.
- SECURITY.md: vulnerability reporting guidance.
- examples/structural-report-demo: canonical demo workspace.
- docs/agents/quickstart.md: agent/operator quickstart.
- AGENTS.md: repo-local agent instructions and source-of-truth order.
The current open source wedge is structural calculation and engineering report generation, while the public category is broader: an AI-native engineering workspace. The near-term focus is:
- a clear example workspace
- a reproducible agent demo
- fast local installation
- real pilot users with repeatable reporting workflows
Inspyro is released under the Apache License 2.0.