Skip to content

HUE

A calm, local-first workspace for human-directed AI work—not one all-knowing agent.
SPEC · reviewableTBI · not implemented24 TBD decisions

One workspace. Clear boundaries. Replaceable intelligence.

Section titled “One workspace. Clear boundaries. Replaceable intelligence.”

HUE brings finishable Projects, ongoing Areas, reusable Resources, independent Sessions, durable Tasks, user-owned knowledge and temporary specialist workers into one coherent interface. The workspace owns the context and state; Hermes, OpenCode, model providers and agent runtimes remain replaceable infrastructure.

9milestones
54canonical issues
18specification chapters
15system diagrams

Product experience

Understand the frustrations HUE removes, the Project/Area/Session mental model, core journeys and every major interface state.

Explore the experience →

System design

Review the local-first control plane, layered context, bounded orchestration, worker adapters, permissions and recovery model.

Inspect the architecture →

Interactive specification

Walk through Home, Inbox, Notifications, Projects, Areas, Sessions, execution, approvals, knowledge and roadmap screens.

Open the prototype →

Implementation plan

Follow the canonical milestone sequence, dependency graph and 54 autonomous-agent-ready implementation briefs.

Review the roadmap →

Every capability is marked TBI until working code and evidence exist. Every unresolved material choice is marked TBD. The documentation is the product contract: implementation should realize it, and any intentional product change must update it visibly.