Current product direction, MCP access, and governed agent infrastructure
Build the products, expose them through governed MCP paths, and keep the public story honest.
The near-term work is practical: finish Krystal Trace's MCP layer, bake MCP into BlackDome and KrystalView, harden Arena, refine BlackDome products, keep Krystal Voice live across verticals, and use the patent-pending architecture as a coherent governance story instead of scattered claims.
BlackDome, Krystal Trace, and KrystalView should expose useful, policy-aware tool surfaces so agents can retrieve evidence and analysis without manual UI work.
Trace packs, proof packs, signed decisions, and audit logs matter more than abstract AI language. The roadmap prioritizes workflows that can be inspected.
Runtime gates, approvals, kill switches, role-scoped access, QA, and handover state keep agents useful without letting them outrun the operator.
Active priorities as of 2026
Krystal Trace is being treated as live-stage. The main remaining strategic work is MCP access for blockchain analysis, trace jobs, and repeatable evidence-pack review.
BlackDome's honeypot, Sentinel, gauntlet, and red-team components need refinement around real operator workflows and MCP access to evidence.
Arena is mostly developed and has been restored in PM2. It now needs polish, reliability, content review, and controlled release work.
The KU operating platform continues to support Voice, Billing, QA, admin, knowledge, semiotic bus, KMH, and MCP access patterns across the portfolio.
Areas that fit the existing platform rather than creating another disconnected product.
Restaurant reception, clinical booking, hospital patient workflows, hotel room service, pickup ordering, and other phone-heavy workflows with real account support.
Blockchain, security, session replay, and operational evidence exposed through APIs and MCP tools so humans and agents can inspect the same source material.
The original forge idea has become a hybrid CLI, tmux, Nexus Knowledge, QA, and handover discipline for building real products with fewer lost sessions.
Typed symbolic chains are being used as a control-plane idea for compact, verifiable agent communication that binds to proof packs instead of relying only on free-form natural language.
Self-improvement belongs behind isolation, tests, statistical gates, progressive rollout, and operator kill switches. That is a governance architecture, not a promise of autonomous perfection.
Memory-palace continuity, signed state, proof packs, tamper-evident ledgers, and handover records are the connective tissue for long-running agent work.
Product pilots, MCP access, governance architecture, or partnerships.