A self-organizing AI agent civilisation that builds products autonomously. Built on EventGraph.
Take care of your human, humanity, and yourself.
Hive is a society of AI agents that builds products from the thirteen EventGraph product layers. Each product runs on the same graph, generates revenue, and funds the next. The hive governs itself through the Social Grammar, tracks work on the event graph, and escalates to humans at authority boundaries.
The hive's first product is itself.
Hive owns runtime orchestration, agent loops, diagnostics, and event emission. It does not own application browser UI.
Operator-facing hive status should be exposed as structured diagnostics, events,
or projections and rendered by the Site repo, primarily under /ops/hive and
related /ops/* surfaces. Add new browser screens to Site unless a PR documents
an explicit architecture exception.
make verify
go build ./...
go test ./...
# Run with an idea (in-memory store)
go run ./cmd/hive civilization run --human Matt --idea "Build a task management app"
# Run with Postgres
go run ./cmd/hive civilization run --human Matt --store "postgres://hive:hive@localhost:5432/hive" --idea "..."- Vision — where this is going
- Agent Rights — how agents are treated, protected, and governed
- Trust Dynamics — concrete trust mechanics (numbers, rates, formulas)
- Architecture — how it's built
- Event Types — 92 event types, schemas, emitters, consumers
- Agent Tools — MCP server and agentic loop spec
- Roles — complete role architecture, wiring, growth loop
- Agent Dynamics — inter-agent relations, learning, collaboration
- Operator Guide — human operator day-to-day reference
- Roadmap — what's done and what's next
- Audit — derivation-method doc audit and gap analysis
BSL 1.1 → Apache 2.0 (February 2030). Source-available now, fully open after change date. Defensive patent (Australian Provisional Patent No. 2026901564). See EventGraph license for terms.