# Changelog

## 0.1.0-rc.1 (2026-10-30)

- `use Wardrik`: one module in your supervision tree that runs agents, keeps them in a
  `WardrikStore`, and holds them to a deployment-wide policy.
- Agents defined as data: model, instructions, tools, policy and triggers. Ask them, stream their
  answers, or wait for a `Wardrik.Reply`.
- Policies held at once by the deployment and each agent: rules that allow, ask or deny, budgets
  per agent, and ceilings per org.
- Approvals, for the people you choose to decide the calls agents wait on.
- Scopes that act as a person, with grants, roles, API tokens and your own authorizer.
- MCP servers Wardrik starts with, and servers an org or person adds as it runs. Credentials sealed
  at rest, and sign-ins that give agents people's accounts.
- Every operation as a function that takes and answers structs. Events and an audit to follow into
  your app.
- Guides for each, and Livebook notebooks that show them in action.