Cranley docs
Cranley is one place where an agent assembles the tools a person’s need requires and checks real buildings, inside and out. These docs cover how to run it, the ideas it is built on, and a reference for every part.
The code name threshold stays on the Python package, the command line and the MCP server. Where a command says threshold, that is Cranley.
Start here
Section titled “Start here”- Getting started: run the server and the web app, with your own keys or with none.
- FAQ: what it can and cannot tell you.
Concepts
Section titled “Concepts”- The workspace: one document that the agent and the person both edit.
- Tools: a manifest and a handler.
- Personas: presets of tools and thresholds for a kind of user.
- Findings and reliability: where every number comes from.
- The agent: the composer and the investigator.
Reference
Section titled “Reference”- Tool reference: one page per tool, generated from its manifest and from a real run on a made-up test scan.
- MCP server and CLI: every tool from Claude Code, Claude Desktop, Cursor or a terminal.
- HTTP API: generated from the server’s OpenAPI schema.
- Result format, personas and map layers.
Project
Section titled “Project”How these docs stay true
Section titled “How these docs stay true”The tool pages, the API pages, the personas, the map layers, the result format and the MCP page are written by site/scripts/generate.py from the code. When a tool or a route changes, running the generator again changes the page. The pages written by hand are this one, getting started, the concepts, architecture, data sources, the phone page and the FAQ.