Documentation
Ductile AI documentation
How to set up and run Ductile AI for your organisation, and how to reach it from your own code.
Start here
Free plan guide
Start a job, bring your own model, and how credits work on Free.
Free plan guideStart a jobBring your own modelCredits and limitsPro and Team guide
Development and Production, and moving a job between them with GitHub.
Pro and Team guideDevelopment and ProductionFrom Development to ProductionAdmin guide
People and roles, single sign-on, the AI gateway, your own runners, runs, logs and transcripts.
Admin guideMembers and rolesSingle sign-onYour own runnersBuild with code
The console as an MCP server, the REST API, and what to use from JavaScript and Python.
MCP serverREST APISDKThe platform
What Ductile AI is, its architecture, and how a job gets done.
Platform overviewArchitectureBring your own model and podLibraries
Processes ready to adapt, and every formula with examples.
Process libraryFormula referenceAppsWhere things live
| Console | https://console.ductileai.com |
|---|---|
| MCP server | POST /mcp, JSON-RPC 2.0, protocol 2026-07-28 |
| REST API | /api/v1/console/… |
| Authentication | a personal access token, Authorization: Bearer dpt_…, minted in the console under Settings › Security › Access tokens |
| Tenancy | every call acts inside the token owner's organisation, with that person's permissions; a token can do nothing its owner cannot |
Sample data first. Every app action ships sample records, and a process runs on those before anything real is connected; the MCP tools and the API honour that.
Last updated October 29, 2026