Build software you can reason about

Septa is a new way to build with Rust. AI agents write within a carefully defined Rust profile. You define the intent, shape the architecture and inspect the result through durable actors with typed contracts, explicit effects and precisely scoped capabilities.

You can build applications, libraries, databases and frameworks. The same foundations support a business service that waits for approval, an agent that invokes tools, and a reusable library consumed by another application.

Choose a first result

I want to…	Start with…	What you will inspect
Run something small	Your first durable actor	State that survives the process that changed it
Build something reusable	A library and its consumer	A library change and the tests that exercise it
Guide a coding agent	Code and evidence	Source context, exact identities and test applicability
Understand the architecture	The system model	Who owns state, who communicates and where effects happen

The tutorials use the current executable name, effectful. Septa is the product name; command names and Rust imports shown in examples preserve the working interface.

Start from what you know

• Rust: a supported language profile, a different execution model and explicit native integration.
• Elixir or Erlang: familiar state owners and messages, with durability and authority made explicit.
• Functional programming: predictable decisions, visible interactions and recorded inputs.
• Building on your own: a small domain slice you can guide, run and understand.
• Working in a team: contracts and evidence that give reviews a shared basis.
• Evaluating for an organisation: authority, trust boundaries, recovery and qualification.

These paths lead to the same platform. Choose the one that gives you the most useful starting vocabulary.

Learn by intent

Tutorials guide you to a working result. How-to guides help you finish a particular task. Explanations develop the mental model. Reference describes exact terms, command forms and guarantees.

You do not need to understand the storage format or every capability rule before running the counter. Begin with one result, make a small change, then follow the explanation of what happened.

When you are assessing a system for real use, read the guarantees and boundaries. Each promise has a scope: a type check, a replay and a test against a live service answer different questions.