Project author

Oleg Sakharov

Revit API / AI engineer

Revit API, AI and MCP in one engineering loop

I build tools that solve problems instead of repeating the lecturer's steps. The same approach — AI agents, MCP and an engineering process — transfers to any other task, not only Revit.

Revit plugins and AI agents that solve engineering problems

Get access What you can build

About

I build tools for Revit and the AI agents that drive them: from a first ExternalCommand to a built, tested plugin handed over to other people.

I run AI × Revit Community — a starter repository, an MCP server, a course and project walkthroughs. Every example follows the same route the landing page describes: task, plan with AI, development, verification, build, release.

Revit is the first domain and the practice ground here: the cost of a mistake is high, which is exactly why setting a task, checking the result and owning the change becomes visible. The skill itself transfers to spreadsheets, reports, integrations and internal utilities.

What you can learn

Revit plugins

From a first ExternalCommand to a published repository.

Revit API
WPF and form applications
MCP server for Revit
Packaging and releasing a plugin

AI agents

Agents that write code and work with tools.

Prompt engineering
Tool-calling
RAG over your own content
Agent web search

MCP (Model Context Protocol)

Connecting tools to models over a standard.

Server and tools
JSON schema contract
stdio and HTTP transports
Key handling and safety

Beyond Revit

The same stack for any automation task.

Tables and documents processing
Git and code review
Workflow automation
Internal utilities

Experience

2024 — present
Author and lead engineer — AI × Revit Community
A plugin starter repository, an MCP server, a course and project walkthroughs: the full path from an idea to a released tool.
- One tool contract and a registry, so a new command plugs in without touching the core
- A calculation core with no Revit dependency and tests on top of it
- Agent orchestration: planner, developer, tester, reviewer, releaser

Projects

RevitAiTools

active

A C# plugin and an MCP server that drive Revit from an AI client: tools are called by the model and executed on the Revit thread.

Designed the single tool contract, the application layers and the route from a wish to a working ribbon command, and turned them into a repeatable procedure.

AI handles the plan, generation, review and tests; a person sets the task, checks the result and decides on the release.

Builds for Revit 2022–2027, an installation guide, built-in diagnostics and repository updates.

C#MCPRevit API.NET

AI × Revit Community

active

The environment around the repository: a course, an engineering library, a digest and a community of engineers.

Brought the repository, the course and the community into one learning route where every lesson sits on the same files as the project.

AI agents are part of the process: context, reuse, review and verification instead of manual trial and error.

A working site with the course, the digest and an admin panel where page copy is edited without a deploy.

TypeScriptNext.jsPostgreSQLMCP

Publications

Materials

Ship your own plugin
Set up an AI agent
Build an MCP server
Run a project in Git with code review

AI × Revit course

course

From the first task to a released tool: requirements, implementation, tests, agents, release.

AI × Revit digest

digest

Walkthroughs of tools and practical uses of AI in engineering work.

Contact

Questions about the course, the repository and working together — through community support.

Write

Project materials and task walkthroughs live in the digest and the site articles.