Skip to content

Roadmap

Michael Elliott edited this page Apr 4, 2026 · 3 revisions

Roadmap

Release History

Phase 1 — Foundation

21+ critical bug fixes, dead code wiring, initial test suite (99 tests).

Phase 2 — Production Grade

Provider router, retry logic, gateway hardening, mesh networking, 10 additional providers, YAML/JS custom skills.

Phase 2.5 — OpenClaw-Inspired

Slash commands, session overrides, auth profile rotation, model allowlisting, provider normalization.

Phase 3 — Full Feature Build (v2026.4.32)

New Skills (8): GitHub, Email, Computer Use, Image Generation, PDF Parsing, Calendar, Smart Home, Data Analysis. New Channels (4): WhatsApp, Matrix, Signal, MS Teams. Infrastructure: Real cron execution, morning briefing, persistent webhooks, dashboard overhaul. Stats: 49+ skills, 57+ tools, 9 channels, 14 providers, 99 tests.

Phase 4–6 — RAG, Streaming, Adaptive Teaching (v2026.7.0)

RAG with FTS5 + embeddings, SSE + WebSocket token streaming, adaptive teaching system.

Phase 7–9 — ContextEngine, Metrics, Deep Research (v2026.8.0)

ContextEngine plugin system, Prometheus metrics, 30 OpenAI-compatible providers (34 total), 15 channels, fallback model chains with self-heal, deep research agent.

Phase 10 — LiveKit Voice (v2026.9.0)

LiveKit WebRTC voice integration replacing custom PCM pipeline. Real-time voice with Orpheus TTS.

Phase 11 — MCP Server Mode (v2026.9.1)

MCP Server mode exposing TITAN tools to other agents via JSON-RPC 2.0 (stdio + HTTP transport).

Phase 12–14 — Mission Control v2 (v2026.10.0)

React 19 SPA with Vite + Tailwind CSS 4, voice health endpoint, distributed setup, THIRD_PARTY_NOTICES.md. 17 admin panels including Mesh, Learning, Autopilot, Security, Workflows, MemoryGraph.

Phase 15–16 — Browser & CAPTCHA (v2026.10.5)

Shared Playwright browser pool, CapSolver CAPTCHA solving (reCAPTCHA, hCaptcha, Turnstile), browser automation skills.

Phase 17–18 — Self-Improvement & Training (v2026.10.10)

LLM-as-judge evaluation, autoresearch, self-improvement system. Local model fine-tuning via LoRA (unsloth) with GGUF export to Ollama. Dual training pipelines (Tool Router + Main Agent).

Phase 19–20 — Teams, Vault, Audit (v2026.10.15)

Team Mode with RBAC (owner/admin/operator/viewer), encrypted secrets vault (AES-256-GCM), HMAC-SHA256 JSONL audit logging.

Phase 21 — Autopilot & Goals (v2026.10.18)

Cron-based autopilot (goals/checklist/self-improve modes), goal management with tracking, self-initiative, sub-agent orchestration. Claude Code agent teams (27 agents, hierarchical org).

Phase 22 — Polish & Scale (v2026.10.22)

Skill scaffolding CLI (JS/TS/YAML templates), one-line installer, cloud deploy support (Railway, Render, Replit), graph memory with entity extraction, income automation skills, X/Twitter integration. ~149 tools across 91 skills.

v1.0.0 — Semantic Versioning Migration

Migrated from date-based versioning (2026.10.XX) to proper semantic versioning. Established 5-file version bump protocol (package.json, constants.ts, core.test.ts, mission-control.test.ts, CHANGELOG.md). Command Post governance with budget enforcement, ancestry validation, stale agent detection. 135+ Playwright E2E tests added.

v1.1.0 — Command Post & Type Safety

Command Post governance hardening, type safety fixes across codebase.

v1.1.1 — CI/Test Infrastructure (Current)

Configurable rate limiter for gateway, GPU auto-tune fix for CI, realistic coverage thresholds. 4,655 tests across 154 files.


Current Stats (v1.1.1)

Metric Value
Version 1.1.1
LLM Providers 34 (4 native + 30 OpenAI-compatible)
Channels 15
Tools ~149 across 91 skills
Tests 4,655 across 154 files
Default Model anthropic/claude-sonnet-4-20250514

Planned Features

  • Marketplace — Skill store with publish/install/ratings
  • Mobile App — React Native companion
  • Workflow Builder — Visual recipe editor with drag-and-drop
  • Multi-Modal Agents — Native image/audio/video processing pipelines
  • Federated Mesh — Cross-network mesh with relay nodes
  • Plugin Ecosystem — Third-party ContextEngine plugins
  • Advanced RAG — Hybrid search with reranking, chunk optimization
  • Agent Memory Sharing — Cross-agent memory synchronization via mesh

Clone this wiki locally