# Providers & Presets
## The runtime contract for your agents

Providers and presets define how Valdr connects to AI models and how agents behave at launch time. A **provider** registers a model backend (Claude, Codex, Ollama, etc.) with its available models. A **preset** pairs a provider with specific runtime configuration — model, permissions, worktree policy, and environment references — into a reusable one-click launch target.

Together they form an explicit contract: changing a model doesn't silently change what tools an agent can access, whether it gets a worktree, or what API keys it needs. **The configuration is inspectable, exportable, and shareable.**

### Why this matters

Swapping models without a configuration layer means:
- Permissions and tool access drift silently
- Worktree isolation gets inconsistent
- API keys get hard-coded or forgotten
- Every team member configures things differently

Providers and presets make the contract explicit. Configure once, export, and every machine runs the same setup.

Use this section to:

- Understand the JSON format for providers and presets
- See real examples from a working Valdr installation
- Learn which fields to customize and which to leave alone

{{< cards >}}
  {{< card link="./overview/" title="Provider Configuration" subtitle="JSON format and field reference for providers" icon="sparkles" >}}
  {{< card link="./presets/" title="Preset Configuration" subtitle="JSON format and field reference for agent presets" icon="code" >}}
  {{< card link="./format/" title="Customization Guide" subtitle="What to change, what to lock down" icon="cube" >}}
{{< /cards >}}