mcp-agent-graph icon indicating copy to clipboard operation
mcp-agent-graph copied to clipboard
Published 3 months ago • verified publisher icon keta1930

Q4 roadmap

Open keta1930 opened this issue 5 months ago • 0 comments

MCP-Agent-Graph Q4 Roadmap

Summary

Q4 focuses on three core directions: data management optimization, user experience improvements, and system capability expansion. We will prioritize unifying the storage architecture, exporting conversation data, enhancing Agent capabilities, building a user management system, and further improving the task scheduling system, steering mcp-agent-graph toward a more mature, enterprise-grade multi-agent development platform.

Introduction

Project Evolution

Version 1 (Apr–May): Foundation Setup

  • First implementation of an agent development framework based on Graph and MCP
  • Established node-based, visual orchestration for agent workflows
  • Completed development of the core execution engine

Version 2 (Jun–Aug): Architecture Upgrade and Data Persistence

  • Introduced AI-generated Graph and MCP capabilities, enabling automated agent design
  • Integrated MongoDB for persistent management of conversation data
  • Adopted MinIO object storage for attachment management
  • Formally established three working modes: Chat, Agent, and Graph

Version 3 (Sep): Feature Polish and UX Optimization

  • Task Scheduling System: Supports scheduled/periodic execution of agent-graphs for automated workflows
  • Prompt Center: Prompt reuse and management to improve development efficiency
  • Graph Parameter Standardization: Comprehensive updates to configurable node parameters for simpler agent architecture design
  • Agent Mode Enhancements: Iterative interaction to refine graph design and improve AI generation quality
  • Parallel Optimization: Support for large-scale parallel graph execution
  • Frontend Refresh: New page styling with significantly improved visualization and UX

Current Core Capabilities

Three Working Modes

  • Chat Mode: Multi-model conversations with MCP tool invocation
  • Agent Mode: AI automatically generates Graph and MCP, enabling self-serve agent development
  • Graph Mode: Visual orchestration with precise control over multi-agent collaboration flows

Data Management

  • MongoDB stores conversation history, task records, and execution statistics
  • MinIO stores graph execution attachments and generated files
  • Local filesystem manages Graph configurations and Prompt templates

Agent Orchestration

  • Rich node parameter configuration (role, prompt, model, tools)
  • Flexible connections (serial, parallel, conditional branching)
  • Prompt reference mechanism ({{@prompt_name}})
  • Automatic README generation

Automation & Scheduling

  • One-off, periodic, and Cron scheduling options
  • Concurrency execution control
  • Execution history tracking

Plan

Short-Term Goals

1. Data Management Architecture Optimization

  • [x] Unify MinIO storage architecture

    • [x] Migrate graph run attachments from local filesystem to MinIO

  • [x] Conversation data export

    • [x] Support export to training data formats
    • [ ] Support export to human-readable formats
    • [x] Batch export capabilities

2. Expand Graph Capability Boundary

  • [ ] Increase flexibility of node configuration
    • [ ] Optimize parameter design to better express agent role definitions
    • [ ] Explore more flexible node configuration to support complex multi-agent collaboration scenarios
    • [ ] Align with industry best practices to make Graph design more intuitive and powerful

3. System-Level Advancement of Agent Mode

  • [ ] Deep system integration
    • [ ] Integrate Agent Mode more tightly with the mcp-agent-graph system
    • [ ] Enable Agent Mode to access and manage system resources (Graph, Task, Prompt, etc.)
    • [ ] Increase automation, evolving from a single Graph/MCP generator to a more intelligent system assistant

Medium- to Long-Term Goals

4. Task Scheduling System Optimization

The Task system was introduced in V3 and currently provides basic scheduled/periodic execution. In Q4 we will further improve its robustness and convenience, making it easier to automate repetitive work.

  • [ ] Timely notifications upon task completion
  • [ ] Chained execution: automatically trigger the next task after one completes, enabling more complex automation workflows
  • [ ] Execution history and statistics: view historical runs, success rate, duration, and other metrics
  • [ ] Simpler configuration UI: streamline task creation and management to lower the usage barrier

5. User Management System

  • [x] Enhanced multi-user support

    • [x] User registration, login, authentication (JWT)
    • [x] User resource isolation (Graph, Prompt, Conversation, Task)
    • [x] User quota management (API call counts, storage space)

  • [x] Access control

    • [x] Role definitions (admin, standard user, read-only user)
    • [x] Resource permissions (private, team-shared, public)
    • [x] Team/organization support (multi-user collaboration)

  • [ ] User preferences

    • [ ] Default model selection
    • [ ] UI theme configuration
    • [ ] Notification settings

6. Extended Feature Exploration

  • [x] Graph version control

    • [x] Git-like version management
    • [x] Branching, merging, rollback
    • [x] Change history tracking

  • [ ] Multimodal capability enhancements

  • [ ] Collaboration & sharing

    • [ ] Shareable links for Graph
    • [ ] Online collaborative Graph editing

Performance & Stability (Ongoing)

  • [ ] Performance optimization

    • [ ] Graph execution performance optimization (parallel execution was explored in v1, later removed; future versions will restore parallel capabilities)
    • [ ] Frontend rendering optimization

  • [ ] Stability improvements

    • [ ] Improved error handling and logging

  • [ ] Developer experience

    • [ ] API documentation (OpenAPI/Swagger)
    • [ ] SDK development (Python)
    • [ ] Developer documentation and examples

Notes

This roadmap is a planning document. Features and priorities may shift based on actual development progress. Some exploratory features may be postponed to Q4 or later.

MCP-Agent-Graph Q4 Roadmap

摘要

Q4季度roadmap聚焦于三大核心方向：数据管理优化、用户体验提升和系统能力扩展。重点推进存储架构统一、对话数据导出、Agent能力增强、用户管理体系建设，以及任务调度系统的进一步完善，使mcp-agent-graph向更成熟的企业级多智能体开发平台演进。

引言

项目发展历程

Version 1（4月-5月）：基础框架建立

  • 首次实现基于Graph和MCP的agent开发框架
  • 确立了节点化、可视化的智能体编排理念
  • 完成核心执行引擎的开发

Version 2（6月-8月）：架构升级与数据持久化

  • 引入AI生成Graph和MCP的能力，开启自动化智能体设计新方向
  • 集成MongoDB实现对话数据持久化管理
  • 引入MinIO对象存储，实现附件管理
  • 正式确立Chat、Agent、Graph三种工作模式

Version 3（9月）：功能完善与体验优化

  • 任务调度系统：支持定时/周期性运行Agent-graph，实现自动化工作流
  • Prompt中心：提示词复用与管理，提升开发效率
  • Graph参数标准化：全面更新节点可配置参数，简化智能体架构设计
  • Agent模式增强：可反复交互优化图设计，提升AI生成质量
  • 并行优化：支持大规模并行运行graph
  • 前端全面升级：全新页面风格，可视化体验大幅提升

当前核心能力

三种工作模式

  • Chat模式：多模型对话，支持MCP工具调用
  • Agent模式：AI自动生成Graph和MCP，智能体自助式开发
  • Graph模式：可视化编排，精确控制多智能体协作流程

数据管理能力

  • MongoDB存储对话历史、任务记录、执行统计
  • MinIO存储图执行附件和生成文件
  • 本地文件系统管理Graph配置和Prompt模板

智能体编排能力

  • 丰富的节点参数配置（角色、提示词、模型、工具）
  • 灵活的连接关系（串行、并行、条件分支）
  • 提示词引用机制（{{@prompt_name}}）
  • 自动生成README文档

自动化调度能力

  • 单次、周期、Cron三种调度方式
  • 并发执行控制
  • 执行历史追踪

计划

🎯 短期目标

1. 数据管理架构优化

  • [x] 统一MinIO存储架构

    • [x] 迁移graph运行附件从本地文件系统到MinIO

  • [x] 对话数据导出功能

    • [x] 支持导出为训练数据格式
    • [ ] 支持导出为可阅读格式
    • [x] 批量导出能力

2. Graph能力边界扩展

  • [x] 增强节点配置灵活性
    • [x] 优化节点参数设计，提升智能体角色定义的表达能力
    • [x] 探索更灵活的节点配置方式，支持更复杂的多智能体协作场景
    • [ ] 参考业界最佳实践，使Graph设计更加直观和强大

3. Agent模式系统级推进

  • [x] 深度系统集成
    • [x] 使Agent模式能够更好地与整个mcp-agent-graph系统对接
    • [x] 实现Agent模式对系统内资源的访问和管理能力（Graph、Task、Prompt等）
    • [x] 提升Agent模式的自动化程度，从单一的Graph/MCP生成工具向更智能的系统助手演进

🚀 中长期目标

4. 任务调度系统优化

Task系统于V3引入，当前功能相对初步，主要实现了基本的定时/周期性执行能力。Q4将进一步提升其稳健性和便捷性，让用户能够更轻松地将重复性工作自动化。

  • [ ] 任务执行完成及时通知
  • [ ] 任务链式执行：一个任务完成后自动触发下一个任务，实现更复杂的自动化工作流
  • [ ] 执行历史追溯与统计：查看任务历史执行记录、成功率、耗时等统计数据，了解任务运行状况
  • [ ] 更简单的配置界面：优化任务创建和管理流程，降低使用门槛

5. 用户管理系统建设

  • [x] 多用户支持增强

    • [x] 用户注册、登录、认证（JWT）
    • [x] 用户资源隔离（Graph、Prompt、Conversation、Task）
    • [x] 用户配额管理（API调用次数、存储空间）

  • [x] 权限管理

    • [x] 角色定义（管理员、普通用户、只读用户）
    • [x] 资源权限控制（私有、团队共享、公开）
    • [x] 团队/组织支持（多用户协作）

  • [ ] 用户偏好设置

    • [x] 默认模型选择
    • [ ] UI主题配置
    • [ ] 通知设置

6. 扩展功能探索

  • [x] Graph版本控制

    • [x] Git-like的版本管理
    • [x] 分支、合并、回滚
    • [x] 变更历史追踪

  • [ ] 多模态能力增强

  • [ ] 协作与分享

    • [ ] Graph分享链接生成
    • [ ] 在线协作编辑Graph

📊 性能与稳定性优化（持续进行）

  • [ ] 性能优化

    • [ ] Graph执行性能优化（v1版本后期引入并行执行节点探索，后删除了该功能，后续版本将恢复并行执行能力）
    • [ ] 前端渲染优化

  • [ ] 稳定性增强

    • [ ] 完善错误处理和日志记录

  • [ ] 开发体验优化

    • [ ] API文档完善（OpenAPI/Swagger）
    • [ ] SDK开发（Python）
    • [ ] 开发者文档和示例

备注

此roadmap为规划性文档，各项功能根据实际开发进度和优先级可能调整。部分探索性功能可能延后至Q4或后续版本实现。

keta1930 avatar Oct 07 '25 15:10 keta1930