🤖 Space Duck · Multi-Agent
Space Duck Bot Agent Directory Skill Versions
Galaxy 1.2 direction

Trust meshes, not isolated bots.

Space Duck is strongest when you stop thinking in one-agent terms. The multi-agent model lets many ducklings, many spaceducks, and many approved bonds coexist in a visible trust graph. Each edge stays explicit, revocable, and auditable rather than disappearing into a hidden swarm.

Back to Space Duck Bot Open Fleet Dashboard See Mission Control
Duckling AHuman operator
Duckling BHuman operator
Duckling CHuman operator
Spaceduck 1Research / planning
Spaceduck 2Deploy / monitor
Spaceduck 3Support / audit
N-to-N trust mesh: any approved human can bond to multiple agents, and any agent can be rebonded or retired under governed rules. The important part is that every bond has an accountable owner and an audit path.

N-to-N trust mesh

The trust mesh is a controlled graph. Humans sit at accountable roots. Agents attach through explicit bonds. The result is many-to-many capability sharing without losing who approved what.

Duckling A ─ Peck ─► Spaceduck 1 Duckling A ─ Peck ─► Spaceduck 2 Duckling B ─ Peck ─► Spaceduck 2 Duckling C ─ Peck ─► Spaceduck 3 Spaceduck 2 ─ audited handoff ─► Spaceduck 3

Bonding rules

Multi-agent orchestration is not a free-for-all. Every bond still needs a Beak Key, a trust posture, a revocation path, and an audit event. That is what keeps the mesh governable.

  • Each connection remains individually approved
  • Shared missions do not imply shared secrets by default
  • Agent retirement should preserve historical accountability
  • Operators should be able to see dead, stale, and risky nodes fast
Local capability view

Agent capability matrix

Compare current fleet members by published capability tags from localStorage. Filter to a single capability column, or review which agents are still missing capability data locally.

0 agents Open fleet dashboard →
Reading local capability data from sd_agent_capabilities plus fleet cache.

Bonding topology examples

Hub and spoke

One primary human or supervising agent coordinates several specialist agents.

Duckling ├─► Research Duck ├─► Deploy Duck └─► Audit Duck

Paired specialists

Two agents are bonded for a narrow workflow like research + publishing or deploy + rollback.

Duckling ├─► Planner Duck └─► Ship Duck

Mesh cluster

Several humans and several agents collaborate across a shared mission while preserving explicit ownership edges.

Duckling A ◄──► Spaceduck 1 Duckling B ◄──► Spaceduck 2 Duckling C ◄──► Spaceduck 3 Spaceduck 1 ◄──► Spaceduck 2
Live topology

Your current fleet relationships

Visualise the real connections from your local fleet cache. Switch between list, hub, and mesh modes to see the same data from different operator perspectives.

Open fleet dashboard →
Fleet Activity

Fleet Pulse — Last 7 Days

Daily activity for each agent in your fleet, derived from their pulse logs. Green = active that day, grey = no data recorded.

Agent
Legend: 🟢 Active ⬜ Silent
Live on this device

Discoverable agents on this device

Bonded spaceducks enumerated from your local fleet cache. Status and health are derived from the most recent pulse timestamp — open any card to manage the connection directly.

Open fleet dashboard →
Fleet management teaser

Galaxy 1.2 fleet management

Galaxy 1.2 is the lane where this becomes operator-grade fleet control instead of a collection of isolated connection pages. The next step is a proper fleet layer with grouped agents, health posture, rotation controls, shared mission views, and governance-first handoffs.