The Hive stack has one public map.
Memory is the core. Swarm and Control are extensions. This page keeps the public GitHub story clean and operator-readable.
Memory first, operations second.
The old map made every Hive piece look equal. This hierarchy points users to the foundation first, then the optional operational layers.
HiveMemory
Persistent memory for AI agent teams.
Private agent silos plus a curated shared hive, so sessions resume with less re-explaining and less context-window waste.
- Private agent memory
- Shared hive index
- Curated learnings
- Bootstrap contract
HiveSwarm
Hardware-aware routing for agent work.
Routes work across usable compute lanes, tracks capacity, and degrades honestly when a machine or worker is unavailable.
- Capacity-aware scheduling
- Local + remote workers
- Agent slot visibility
- Graceful offline mode
HiveControl
Owner-consented control for real machines.
Capability-gated actions, proof records, canaries, and approval boundaries for terminal, file, browser, desktop, and worker control.
- Consent boundary
- Relay canaries
- Gateway control
- Recovery notes
Public code needs public-safe proof.
These repos document reusable patterns without exposing private tokens, hostnames, runtime logs, or machine-specific control details.
Owner consent and clear machine boundaries come first.
Live canaries beat stale memory or theoretical configuration.
Transport success is not execution success until output proves it.
Public repos use placeholders and patterns, never private runtime details.