Roadmap

This page tracks what’s shipped and what’s coming. Updated as priorities shift.

Shipped

v0.1 - Foundation

• Message bus architecture (event-driven core)
• Agent loop with memory management
• Tool protocol and registry
• File-based memory system
• Telegram adapter with QR pairing
• Web dashboard (Alpine.js, WebSocket streaming)
• Ollama, OpenAI, Anthropic LLM support

v0.2 - Browser and security

• Playwright browser automation with accessibility tree snapshots
• Element reference system ([ref=N] for clicking, typing)
• Guardian AI safety layer (secondary LLM check before shell commands)
• Audit logging (append-only ~/.mudabbir/audit.jsonl)
• File jail and single-user lock
• Proactive daemon with cron triggers

v0.3 - Channels and integrations

• Discord adapter (slash commands, DM, mention support)
• Slack adapter (Socket Mode, no public URL needed)
• WhatsApp Business Cloud API adapter
• Gmail, Calendar, Drive, Docs integrations via OAuth
• Spotify and Reddit tools
• Web search (Tavily, Brave, Google)
• Image generation (Google Gemini)
• Voice and TTS (OpenAI, ElevenLabs)
• Speech-to-text (Whisper)
• OCR (GPT-4o vision, pytesseract)
• Research tool with source synthesis
• MCP server support (stdio + HTTP)

v0.4 - Production hardening (current)

• Encrypted credential store (Fernet AES, auto-migration from plaintext)
• 7-layer security stack
• Injection scanner (regex + optional LLM deep scan)
• Security audit CLI (mudabbir --security-audit)
• Self-audit daemon (12 daily checks, JSON reports)
• Tool policy engine (profiles: minimal, coding, full)
• Plan mode (approval workflow for tool execution)
• Smart model router (Haiku, Sonnet, Opus based on complexity)
• Session compaction (tiered compression for long conversations)
• Deep Work mode (focused multi-step execution)
• Mission Control (multi-agent orchestration)
• Skills system (runtime agent training via SKILL.md files)
• Channel management UI (configure, start, stop from dashboard)
• Docker support
• 130+ tests passing

Coming next

v0.5 - Desktop and onboarding

• Native installers (.exe for Windows, .dmg for macOS, .AppImage for Linux)
• First-run setup wizard in the web dashboard
• System tray app (start, stop, open dashboard)
• One-click Ollama detection and install
• Auto-update mechanism
• Backup and restore (export config + memory)

v0.6 - WebMCP and browser upgrades

• WebMCP integration for structured website interaction
• Hybrid browsing: WebMCP where available, accessibility tree fallback
• Token usage tracking and cost dashboard
• Model failover and context overflow recovery

Future

• Signal, Matrix, Microsoft Teams, Google Chat adapters
• Docker sandbox for shell command isolation
• Hybrid memory search (vector + keyword)
• Plugin architecture
• Multi-user support
• Session teleportation (cross-device continuity)
• Desktop notifications center

How we prioritize

We ship what unblocks users first. Security and reliability fixes jump the queue. Feature requests from GitHub issues influence what gets built, so if you want something, open an issue.