Settings

Settings is where you wire up the infrastructure behind every agent launch. Define which AI providers are available, create reusable presets that encode exactly how agents should run, and import or export packs to standardize your setup. Configure once, launch confidently every time.

What this page covers

• Agent Presets — Create, edit, export, and import provider configurations
• Providers — Register AI providers and manage model availability
• Valdr Packs — Import and export packs for portable agent configurations

Agent Presets

Agent presets are saved provider configurations. Each preset bundles a provider, model, worktree policy, environment references, and runtime config into a reusable one-click launch target. Instead of re-configuring every agent session from scratch, you pick a preset and go.

All your provider presets in one place — pick a provider, set it up once, reuse everywhere

Preset table columns

Creating a preset

Click the + button in the toolbar to open the preset form. Every field shapes how agents behave when launched with this preset:

Full control over every launch parameter — provider, model, permissions, worktree isolation, and runtime config

For OpenCode, set the desired model (provider/model) and optionally agent (such as build) in Config JSON. Env refs can stay empty, and mcpServers is optional. A preset can also use {} to keep OpenCode’s defaults. See the OpenCode preset example.

For Claude and Codex, presets can set reasoning effort. At launch, Use preset/default inherits that setting; an explicit effort overrides it.

Worktree options

The three checkboxes at the bottom of the preset form control git worktree behavior:

Export and import

Use the Export and Import buttons to move presets between machines or share them with teammates.

• Export saves all presets (or a selection) as a portable file
• Import loads presets from a file, with conflict detection if a key already exists

This makes onboarding fast — set up presets once, export, and every team member imports the same configuration. Combined with Valdr Packs and provider/preset configuration, you can standardize an entire team’s agent infrastructure in minutes.

Providers

The Providers tab is where you register the integrations Valdr can use. Each entry has a type, display name, and enabled status. Model lists and defaults are optional; an OpenCode provider can leave them empty and select models through presets.

Three providers configured — cloud and local models side by side, each with its own model roster

Provider table columns

Adding a provider

Click + to register a new provider:

1. A display name — what you see in dropdowns and tables
2. A type — the integration protocol (claude, codex, opencode, or ollama)
3. Enabled — turn the provider on so it is available for sessions

For OpenCode, leave the API key variable, base URL, default model, and token limit empty. Keep Models as []. Create presets to select the models you want to run; no MCP configuration is required just to add the provider.

Mixing cloud and local providers

Upgrading to 0.3.4: Direct Gemini, OpenAI API, and Anthropic API providers have been removed. Create an OpenCode provider and replacement presets, then update affected agents and workflows. Existing presets are not automatically converted.

Valdr treats all providers equally. Run Claude for complex coding tasks and Ollama for fast local iteration — switch between them by choosing a different preset. This gives you:

• Cost control — use cheaper or free local models for exploratory work
• Offline capability — Ollama presets work without internet
• Comparison — run the same task with different providers and compare session quality in the Agent Sessions view

Export and import

Like presets, providers support Export and Import for portable configuration. Export your provider setup and share the JSON so teammates import identical provider configurations. See Provider Configuration for the full JSON format.

Valdr Packs

The Valdr Packs tab handles import and export of .valdr-pack.tar.gz archives — portable bundles that package agents, prompts, capabilities, and configuration into a single shareable file.

Export your entire agent setup as a portable archive — import it on another machine and be productive immediately

Exporting a pack

1. Select pack keys — choose which packs to include in the export (e.g., valdr)
2. Filter by agent handles (optional) — limit the export to specific agents (e.g., @executor, @reviewer)
3. Click Export Pack to generate the .valdr-pack.tar.gz archive

The exported archive includes all agents, prompts, capabilities, and configuration associated with the selected packs.

Importing a pack

Click Import Pack to load an archive. Valdr runs a conflict-aware preflight check before committing any changes:

• Detects naming conflicts with existing agents, prompts, or capabilities
• Shows what will be added, updated, or skipped
• Requires explicit confirmation before applying changes

This means you can safely import packs without worrying about overwriting your existing setup.

Packs can also carry user tools. Import and export them here, then browse and inspect installed tools in Workflows > Tools.

How settings flow into agent launches

Settings aren’t just configuration — they’re the foundation of every agent session. Here’s how the pieces connect:

1. Register providers in the Providers tab (Claude, Codex, OpenCode, or Ollama)
2. Create presets in Agent Presets that reference those providers with specific models and runtime config
3. Launch agents from Tasks or the Orchestrator using your presets
4. Review results in Agent Sessions — every session records which preset and provider was used

Change a preset once and every future launch inherits the update. No re-configuring, no drift, no “which model was that session using?” questions.

Next steps

Ready to put your settings to work? Launch your first agent from a task and see how presets turn configuration into one-click execution.