OpenTakeoff
opentakeoff.kentucky-ai.com
- Free plan
- Open source
- Windows
- macOS
- Linux
- Web
- API
- Self-hosted
What OpenTakeoff is
OpenTakeoff is a free construction takeoff engine for measuring quantities from building plans. Its browser canvas lets people work with plan sets, while an MCP server connects the tool to AI agents. The browser app opens PDF plan sets, images, and ZIP plan sets, processing them locally without an account or plan upload. Measurements include square feet, square yards, linear feet, and each, along with waste-adjusted order quantities, supporting-materials lists, and roll-goods cuts. Exports include CSV, Excel, and marked PDF files. The MCP server runs through `npx -y opentakeoff-mcp` and supports Claude Code, Claude Desktop, Cursor, and other stdio MCP clients. Agent measurements remain pending until a person approves them in the browser canvas. In the default browser workspace, project files and settings are processed on the device and stored in browser storage. Optional connections include Google Drive and Microsoft 365 cloud storage, configured AI providers, and a service that can send a selected schedule image to Google Gemini. The Apache-2.0 project can be forked for a self-hosted instance. One-Click Area is temporarily gated, and the default MCP build does not register room-detection verbs.
Who it suits
It suits construction professionals measuring plans and developers connecting takeoff work to AI agents. The browser workspace works without an account, while the project can also be forked for self-hosting.
Good to know, and what to watch
- Opens PDF, image, and ZIP plan sets.
- Processes plans locally without an account or upload.
- Exports CSV, Excel, and marked PDF files.
- Agent measurements require human approval.
- Apache-2.0 licensed and available to self-host.
- One-Click Area is temporarily gated.
- Default MCP build lacks room-detection verbs.
- Optional cloud and AI integrations may send selected data.
Plans
Every plan and pricePremium
Free
Free plan
Getting started
- Open the OpenTakeoff website
- Load a PDF, image, or ZIP plan set in the browser
- Measure quantities on the browser canvas
- Export results as CSV, Excel, JSON, PDF, or DXF
- Install the MCP server with `npx -y opentakeoff-mcp` for agent clients
- Optionally connect Google Drive, Microsoft 365, or configured AI providers
Limits to check
The free plan is free and has no account requirement. One-Click Area is temporarily gated, and the default MCP build does not register room-detection verbs; agent measurements remain pending until approved in the browser canvas.
The verdict
OpenTakeoff combines a browser-based takeoff canvas with an MCP server for agent-assisted measurement. Its local default workspace and human approval step are useful distinctions, but some features remain gated or absent from the default MCP build.
Questions people ask
How much does OpenTakeoff cost?
The Free plan costs nothing. It is open-source under Apache-2.0 and requires no account.
Which plan files can it open?
The browser app opens PDF plan sets, PNG and JPEG images, and ZIP plan sets.
What platforms are supported?
Listed platforms are API, Linux, macOS, self-hosted, web, and Windows.
Does it support AI agents?
Yes. Its MCP server runs through `npx -y opentakeoff-mcp` and supports Claude Code, Claude Desktop, Cursor, and other stdio MCP clients.
Are agent measurements applied automatically?
No. Agent measurements stay pending proposals until a person approves them in the browser canvas.
Is OpenTakeoff open source?
Yes. The project is Apache-2.0 licensed, and users can fork it to run their own instance.
What the maker publishes
- Agent integration
- The MCP server runs through `npx -y opentakeoff-mcp` and supports Claude Code, Claude Desktop, Cursor and other stdio MCP clients. source
- Agent review
- Agent measurements remain pending proposals until a person approves them in the browser canvas. source
- Area measurement
- The engine measures areas in square feet or square yards for floors, ceilings and walls. source
- Cloud integrations
- Documentation describes optional Google Drive and Microsoft 365 connections that use the user's sign-in and permissions to synchronize project files. source
- Company founding
- LinkedIn lists Kentucky AI as founded in 2026. source
- Company headquarters
- LinkedIn lists Shelbyville, Kentucky as Kentucky AI's headquarters. source
- Company ownership
- LinkedIn classifies Kentucky AI as privately held and in software development. source
- Company size
- LinkedIn lists Kentucky AI as having 2–10 employees. source
- Count measurement
- Count tools record fixtures, devices and doors as eaches. source
- Data sharing
- Connected AI providers can receive the tool results, images, quantities and project information requested by the user's MCP client. source
- Default feature limit
- One-Click Area is temporarily gated while its flood engine is re-validated; the default MCP build does not register its room-detection verbs. source
- Developer runtime
- The agent guide requires Node 20 or newer and provides an npx configuration without cloning or building. source
- Docker deployment
- The MCP README documents Docker image creation, interactive execution and read-only plan-volume mounts. source
- Documentation depth
- The repository provides separate user and agent manuals, an MCP reference, changelog, feature map, wiki and contributor guidance. source
- Exports
- The app exports CSV, Excel and marked PDF files. source
- Google Drive team mode
- The user guide describes a Google Drive Projects root where PDFs and hidden OpenTakeoff sidecar files live in project folders. source
- Headquarters
- Shelbyville, Kentucky, United States source
- Human review workflow
- Agent-created shapes remain pending and dashed until a person reviews and accepts them. source
- Independent popularity on OpenArena
- OpenArena recorded 124 GitHub stars and 47 forks for the project snapshot shown on October 14, 2026. source
- Independent ranking on OpenArena
- OpenArena ranked OpenTakeoff first among 35 projects in its Tool category. source
- Intended users
- The privacy policy says the service is intended for construction professionals and developers, not children under 13. source
- License in current repository
- The current GitHub README identifies the software as AGPL-3.0-only, while noting earlier commits were Apache-2.0. source
- Limit: computer vision
- Advanced computer-vision models are identified as still being in the lab. source
- Limit: One-Click Area
- The repository says One-Click Area and its related MCP verbs are temporarily gated while the flood engine is revalidated. source
Checked against the maker's pages on 14 Oct 2026.