Roadmap

This page tracks the current state and future direction of MateClaw.

Completed

Agent Engine

• [x] BaseAgent abstract base class (chat / chatStream / execute)
• [x] ReAct Agent (Thought -> Action -> Observation loop)
• [x] DynamicAgent (runtime Agent loaded from database config)
• [x] Plan-and-Execute orchestration (PlanningFactory, Plan / SubPlan models)
• [x] AgentState lifecycle management
• [x] StateGraph Agent architecture (state-graph-based Agent orchestration)

Tool System

• [x] Tool registration mechanism (Spring AI @Tool annotation)
• [x] 12 built-in tools (WebSearch, DateTime, etc.)
• [x] MCP protocol support (stdio / SSE / streamable_http transport)
• [x] ToolGuard tool approval mechanism
• [x] FileGuard file access control
• [x] AuditLog security audit logging
• [x] Tool management UI
• [x] ShellExecuteTool cross-platform support (Windows cmd.exe + timeout fix)

Skill System

• [x] SKILL.md protocol specification
• [x] Skill database storage (mate_skill table)
• [x] Skill install / uninstall / enable / disable
• [x] SkillSecurityService security scanning
• [x] SkillFileAccessPolicy file access control
• [x] SkillDependencyChecker dependency checking
• [x] Skill Market UI
• [x] Skill Workspace filesystem (convention-based directory structure + archive policy)
• [x] Auto-sync bundled skills from classpath on startup

Model Support

• [x] DashScope (Qwen series) integration
• [x] Ollama local model support
• [x] OpenAI-compatible protocol
• [x] Multi-provider management (160+ model configurations)
• [x] LLM built-in search (enable_search / search_strategy)
• [x] Kimi $web_search integration
• [x] Search config UI migration (from config files to System Settings page)

Channel Adapters

• [x] Web (SSE streaming)
• [x] DingTalk bot adapter (Stream long connection + AI Card streaming cards)
• [x] Feishu bot adapter (WebSocket long connection)
• [x] WeChat Work adapter (long connection mode)
• [x] Telegram Bot adapter (default Long-Polling + Typing indicator)
• [x] Discord Bot adapter (JDA Gateway WebSocket)
• [x] QQ channel bot adapter
• [x] WeChat personal adapter (iLink protocol, experimental)
• [x] Channel proxy support (Telegram / Discord http_proxy configuration)
• [x] Session source tracking (IM channel sessions display source icon)

Frontend Console

• [x] Vue 3 + Pinia + TailwindCSS 4 + Element Plus
• [x] ChatConsole chat interface (SSE streaming)
• [x] AgentWorkspace Agent management
• [x] Dark mode support
• [x] Internationalization (Chinese / English)
• [x] Session management

Authentication & Security

• [x] JWT authentication
• [x] JWT sliding-window renewal
• [x] Unified 401/403 error handling
• [x] Spring Security integration
• [x] ToolGuard tool execution guard
• [x] FileGuard file access control rules
• [x] AuditLog security audit logging

Memory System

• [x] Short-term memory (session context window)
• [x] Automatic context compression (compact-trigger-ratio)
• [x] WorkspaceMemoryTool (PROFILE.md / MEMORY.md / daily memory)
• [x] Automatic memory extraction after conversations
• [x] Memory consolidation service (CronJob seed data driven, runs daily at 2:00 AM)

In Progress

Skill Ecosystem (ClawHub)

• [x] ClawHub skill market UI (browse + install)
• [x] Skill Workspace filesystem
• [ ] Skill version management
• [ ] Skill ratings and reviews

MCP Enhancements

• [x] MCP Streamable HTTP transport support
• [x] MCP stdio config hot-reload (no restart required)
• [x] MCP Server dynamic loading and unloading
• [ ] MCP connection pool management
• [ ] MCP health checks and auto-reconnect

Desktop Application

• [x] Electron framework migration (with bundled JRE)
• [x] electron-updater full-package auto-update
• [x] WeChat Work scan-to-authorize popup support
• [ ] System tray integration improvements

v1.0.101

• [x] Drag-and-drop file upload (files and folders)
• [x] Directory reference (Electron local path / Web recursive upload)
• [x] Multi-Agent collaboration (DelegateAgentTool)
• [x] Ollama auto-discovery on startup + 6 pre-configured models
• [x] Zhipu AI model update (GLM-5-Turbo / GLM-5V-Turbo / GLM-5 / GLM-5.1)
• [x] Zhipu connection test fix (completionsPath adaptation)
• [x] Model management page grouping (Local Models / Cloud Models)
• [x] Local provider priority sorting + is_local fix
• [x] File upload limit raised to 100MB
• [x] SpringDoc OpenAPI replaces Knife4j
• [x] Versioned documentation (v1.0.0 / v1.0.101)

Planned

Feature	Priority	Description
Multi-User Support	High	Role-based access control with workspace isolation
RAG Pipeline	High	Retrieval-Augmented Generation with document indexing and vector database
Voice I/O	Medium	Speech-to-text input and text-to-speech responses
Workflow Builder	Medium	Visual drag-and-drop workflow editor for agent pipelines
Observability Dashboard	Medium	Agent tracing, token usage stats, tool call analytics
Image Generation	Low	Integration with image generation models (DALL-E, Stable Diffusion)
Multi-Tenancy	Low	Tenant data isolation and quota management
SSO Integration	Low	LDAP, SAML, OAuth2 single sign-on
Mobile App	Low	Responsive web app or native mobile client

Milestones

Version	Goal	Status
v1.0	Core features stable (Agent, Tools, Skills, Models)	Released
v1.1	8 channel adapters + Memory system + StateGraph + Security enhancements	Released
v1.0.101	Drag-drop upload + Multi-Agent + Ollama auto-discovery + GLM-5 + Model grouping UI	In Development
v1.2	Skill Workspace + Electron desktop app + Channel enhancements	Planned
v2.0	Multi-user + RAG + Workflow builder	Planned

Contributing

We welcome contributions to any roadmap item. If you want to work on a planned feature:

1. Open a GitHub issue to discuss the approach
2. Fork the repository and create a feature branch
3. Submit a pull request following the Contributing Guide

Feedback

Have a feature request or suggestion? Open an issue on GitHub with the enhancement label.