Fleet Plans

A plan is a template configured for a specific mission. It binds a team to a project, channel, and credentials.

Plan Lifecycle

1. Create — Through the Studio UI wizard or the /fleet-plan slash command. The AI-guided wizard walks you through team selection, credential mapping, and channel configuration.
2. Validate — The validate_fleet_plan tool tests external connections (channels, credentials, artifact destinations).
3. Save — The save_fleet_plan tool persists the plan.
4. Activate — Start the plan. This begins polling for work on the configured channel.
5. Execute — Sessions are created for each incoming task.
6. Deactivate — Stop polling.

Plan Configuration

Field	Description
base_fleet_key	Which template to use.
channel_type	How work arrives: chat (manual) or github_issues (automated polling).
channel_config	Channel-specific settings (e.g., GitHub repo, labels to watch).
channel_schedule	Cron expression for polling non-chat channels.
credentials	Mapping of credential names for external services.
artifacts	Where to store outputs.
behavior_overrides	Per-agent instruction overrides.
include_agents	Subset of agents to include from the template.
project_source	Where the project code lives.

GitHub Issues Integration

When channel_type is github_issues, the plan polls a repository for new issues matching configured labels. Each matching issue becomes a fleet session, automatically assigning work to the team.

CLI Commands

# List all plans
astonish fleet list

# View plan details
astonish fleet show <key>

# Start polling
astonish fleet activate <key>

# Stop polling
astonish fleet deactivate <key>

# Check status
astonish fleet status <key>

# Remove a plan
astonish fleet delete <key>