Skip to main content

Seraph Development Status

Legacy note: this page now lives under /legacy. The canonical shipped-status ledger is now docs/implementation/STATUS.md.

Historical note: older lines on this page may still refer to pre-cockpit planning language. Current Seraph is cockpit-only; the village/editor line is retired and removed from the active repo path.

Legend​

  • [x] shipped on develop
  • [ ] not fully shipped on develop
  • In-flight branch work should be tracked in open PRs, not in this file.

Current Snapshot​

  • Seraph is usable today as a local guardian prototype with a real browser UI, observer loop, and action layer.
  • Trust Boundaries, Execution Plane, and Runtime Reliability have strong shipped foundations.
  • The live planning surface is now docs/implementation/ plus docs/research/.
  • No workstream is complete yet.
  • Seraph is not yet a finished product or a production-ready guardian platform.

Currently Active On develop​

  • Runtime Reliability is the current hardening track.
  • The active runtime workstream is not finished yet.

Shipped On develop​

Trust and control​

  • Tool policy modes for safe, balanced, and full
  • MCP policy modes for disabled, approval, and full
  • High-risk approval gates for chat and WebSocket execution
  • Structured audit logging for tool, approval, and runtime events
  • Secret redaction and scoped secret-reference handling

Execution and integrations​

  • Shell execution foundation
  • Browser automation foundation
  • MCP integration and runtime-managed server configuration
  • Auto-discovered built-in native tools and SKILL.md skill loading
  • Visible tool-execution streaming in chat

Runtime and observability​

  • Shared provider-agnostic LLM runtime settings
  • Ordered fallback chains across completion and agent-model paths
  • Health-aware rerouting away from recently failed targets
  • Runtime-path-specific primary model overrides for completion and agent-model paths
  • Runtime-path-specific fallback-chain overrides for completion and agent-model paths
  • First-class local runtime routing for helper, scheduler, core agent, delegation, and connected MCP specialist paths
  • Runtime audit visibility across chat, WebSocket, scheduler, strategist, MCP, observer, embedding, vector-store, soul-file, filesystem, browser, sandbox, and web search flows
  • Deterministic runtime eval harness for fallback, local routing, context-window degradation, MCP specialist routing, embedding/vector-store/soul-file/filesystem boundaries, browser/sandbox/web-search tool, and observer contracts

Product surfaces​

  • Browser-based guardian cockpit UI
  • WebSocket session flow
  • Native macOS observer daemon
  • Proactive delivery inside the current product
  • Soul, memory, goals, strategist, daily briefing, and evening review foundations

Ecosystem foundations​

  • SKILL.md support
  • MCP-powered extension surface
  • Recursive delegation foundations behind a flag

Still To Do On develop​

Runtime Reliability​

  • richer provider selection beyond explicit runtime-path primary and fallback overrides, ordered fallback chains, and cooldown rerouting
  • broader local-model routing into any remaining runtime paths where it makes sense
  • remaining edge observability beyond the already-covered agent, scheduler, observer, and integration paths
  • broader eval coverage beyond deterministic seam checks

Product expansion​

  • native desktop shell, notifications, and external channels
  • deeper guardian world-modeling, learning loops, and stronger intervention quality
  • richer cockpit density, operator visibility, and continuity across the current guardian surfaces
  • stronger workflow composition and extension ergonomics

Workstream View​

  • Workstream 01: Trust Boundaries is only partially complete
  • Workstream 02: Execution Plane is only partially complete
  • Workstream 03: Runtime Reliability is only partially complete
  • Workstream 04: Presence And Reach is only partially complete
  • Workstream 05: Guardian Intelligence is only partially complete
  • Workstream 06: Embodied UX is only partially complete
  • Workstream 07: Ecosystem And Leverage is only partially complete