Key Features
This page maps Wardian's major capabilities to their detailed guides.
Work with Agents Your Way
- Your prompts, classes, skills, automations, and results are saved as files you can inspect instead of being hidden only inside the app.
- Use the desktop app, CLI, Library, Automations, Explorer, Inbox, and Graph with the same saved Wardian data.
- Start with an agent, then save a prompt, improve a class, share a skill, or automate an automation when that helps.
- Share reusable setup with every agent, a class, one agent, a team, a workspace, or an automation run as appropriate.
Related docs:
Monitor Live Agents
- High-density Grid and Dashboard views for live local agent sessions
- Left control rail for spawning, commands, automations, Explorer, and settings
- Right roster with status lights, thought snippets, and agent working sets
- Inbox tab for unread agent completions, important updates, and automation outcomes
- Use Key Concepts to understand the app layout and how its parts connect.
Related docs:
Agent Classes and Reusable Library
- Reusable agent setups with instruction files and assigned skills
- Prompt, skill, class, and automation management from a filesystem-backed library
- Starred prompts for one-click execution from the Command panel
Related docs:
Broadcast and Prompt Injection
- Broadcast text commands to selected agents or all active agents
- Inject starred prompts from the library after flattening into terminal input
Related docs:
Agent-Facing Wardian CLI
- Give agents a textual control surface for inspecting live or persisted Wardian state
- Let agents spawn, clone, pause, resume, kill, wait on, and watch peers through the same backend control layer used by the desktop app
- Send prompts from inline text, standard input, or files; use
wardian askfor bounded peer handoffs with response evidence - List automations, show automation definitions, and start or stop automation runs when the desktop app is available
- Inspect Wardian-managed agent worktrees, teams, and watchlists for automation-friendly coordination without treating the CLI as the primary human UI
Related docs:
Source Control Per Agent
- Git status, staging/unstaging, discard, commit, pull/push
- Inline diff viewing and commit history
- Optional worktree mode for branch isolation
Related docs:
Automation Builder and Runtime
- Visual node-based builder with variable assistant
- Manual, scheduled, and listener-style trigger behaviors
- Scheduled run management and run-time role assignment
- Deterministic Rust automation engine with pulse-driven candidate-node execution, branch/loop/wait control, agent execution modes, run-scoped storage, and live telemetry
Related docs:
Inbox and Completion Triage
- Captures explicit provider turn completions for named agents with their canonical final response, without treating Idle transitions or arbitrary terminal output as a final summary
- Receives explicit agent updates and exceptional manual approval requests through
wardian notify - Records completed and failed automation outcomes for later review
- Persists Inbox completion projections under the Wardian home so unread work survives app restarts
- Supports unread badges, mark-read, clear-read, dismiss, and expandable long summaries
Related docs:
Runtime Shell and Session Policy Controls
- Default shell selection (
auto, discovered shell, or custom executable) - Per-installation regular-agent session persistence policy (
resumevsfresh) - Theme sync support and Gemini patch actions from settings
Related docs:
Provider-Aware Runtime
- Unified runtime model over Antigravity, Claude, Codex, OpenCode, and Pi (Gemini is unmaintained — use Antigravity for Google-model access)
- Provider-specific lifecycle handling for resume/session identity and skill discovery
Related docs: