How it fits together
1
Add context
Documentation belongs to a workspace. On Context, add a source: a repository’s own markdown, or a documentation site through its llms.txt. A Document scan curates the documents of every source into one corpus. See Document scan.
2
Connect a repository
On Code, connect a repository and choose the sources it reads. Flow setup starts in the background. It works out how to build, start and seed the program. See Flow setup.
3
Generate and run flows
Flow generation derives claims from the documents, groups them into flows, and writes a test for each flow. A Flow run executes the tests. See How it works.
The web app
TrueCourse is one web app. Its sidebar holds these pages:Where it runs
Everything durable lives in Postgres. You can run TrueCourse on one machine in local mode, with no sign-in. You can also deploy it for a team, with sign-in. See Installation and Storage.Models
Document scan, Flow setup and Flow generation call a model. Each workspace saves its own provider on Settings › Models: Anthropic API, OpenAI, AWS Bedrock, GitHub Copilot or Google Gemini. A self-hosted instance can instead run every workspace on its operator’s Claude Code login. A Flow run is deterministic and needs no model. See LLM transport.Editions
TrueCourse is open source under the MIT license, except for three enterprise features: document Connections, more repository providers, and more than one workspace. See Enterprise edition. Today GitHub is the only provider that connects, plus local folders when the server runs in local mode. GitLab and Azure DevOps are listed in the app as coming soon.Get started
Quickstart
Run TrueCourse locally and follow your first flows to a result.
Installation
Requirements, environment variables, local mode and hosted mode.
How it works
Claims, flows, milestones and cases, and the jobs that produce them.
Document scan
How the workspace’s documents become one curated corpus.