MassGen Roadmap#
Full Roadmap#
MassGen Roadmap#
Current Version: v0.1.14
Release Schedule: Mondays, Wednesdays, Fridays @ 9am PT
Last Updated: November 19, 2025
This roadmap outlines MassGen’s development priorities for upcoming releases. Each release focuses on specific capabilities with real-world use cases.
👥 Contributors & Contact#
Want to contribute or collaborate on a specific track? Reach out to the track owners below:
Track |
GitHub |
Discord |
|---|---|---|
Tool System Refactoring |
danrui2020 |
|
Multimodal Support |
danrui2020 |
|
General Interoperability |
danrui2020 |
|
RL Integration |
danrui2020, ram2561 |
|
Agent Adapter System |
ericshang. |
|
Framework Streaming |
ericshang. |
|
Irreversible Actions Safety |
zhichengren |
|
Computer Use |
zhichengren |
|
Memory Module |
danrui2020, nickcrispino |
|
Rate Limiting System |
abhimanyuaryan |
|
DSPy Integration |
ram2561 |
|
MassGen Handbook |
crinvo, henry_weiqi |
|
Session Management |
nickcrispino |
|
Automatic MCP Tool Selection |
nickcrispino |
|
Parallel File Operations |
nickcrispino |
|
MassGen Terminal Evaluation |
nickcrispino |
|
Web UI |
justin_zhang |
For general questions, join the #massgen channel on Discord
Release |
Target |
Feature |
Owner |
Use Case |
|---|---|---|---|---|
v0.1.15 |
11/21/25 |
Integrate RL into MassGen |
@qidanrui @praneeth999 |
Reinforcement learning integration for agent optimization and adaptive behavior |
Git Worktrees for Multi-Agent |
@ncrispino |
Use Git worktrees for different agents to enable parallel development workflows |
||
v0.1.16 |
11/24/25 |
Launch Custom Tools in Docker |
@ncrispino |
Enable custom tools to run in isolated Docker containers for security and portability |
Improve Consistency of Memory & Tool Reminders |
@ncrispino |
Enhance memory and tool reminder consistency across agents |
||
v0.1.17 |
11/26/25 |
MassGen Terminal Evaluation |
@ncrispino |
Self-evaluation and improvement of frontend/UI |
Smithery MCP Tools Support |
@ncrispino |
Expand MCP tools access through Smithery integration |
All releases ship on MWF @ 9am PT when ready
📋 v0.1.15 - RL Integration & Multi-Agent Git Workflows#
Features#
1. Integrate RL into MassGen (@qidanrui, @praneeth999)
Issue: #527
Reinforcement learning integration for agent optimization
Adaptive agent behavior based on feedback and outcomes
Reward modeling for multi-agent coordination
Policy optimization for task execution strategies
Learning from past interactions to improve future performance
Integration with existing agent coordination infrastructure
Use Case: Enable agents to learn and improve their performance over time through reinforcement learning, optimizing coordination strategies and task execution based on past successes and failures
2. Git Worktrees for Different Agents (@ncrispino)
Issue: #514
Enable multiple agents to work on different Git worktrees simultaneously
Isolated working directories for parallel agent development
Automatic worktree creation and management
Branch synchronization across agent worktrees
Conflict resolution support for multi-agent workflows
Improved parallelism for multi-agent code development tasks
Use Case: Allow multiple agents to work on different features or branches simultaneously without conflicts, enabling true parallel development workflows
Success Criteria#
✅ RL framework successfully integrates with MassGen architecture
✅ Agents demonstrate learning and improvement over repeated tasks
✅ Reward modeling accurately reflects task success metrics
✅ Policy optimization improves coordination strategies
✅ RL integration maintains system stability and performance
✅ Agents successfully create and manage separate Git worktrees
✅ Multiple agents can work on different branches simultaneously
✅ Worktree cleanup and management works reliably
✅ Branch synchronization maintains code integrity
✅ Performance improvements in multi-agent development scenarios
📋 v0.1.16 - Docker Tools & Expanded Integrations#
Features#
1. Launch Custom Tools in Docker (@ncrispino)
Issue: #510
Enable custom tools to run in isolated Docker containers
Automatic containerization of custom tool execution
Security isolation for untrusted or experimental tools
Improved portability across different environments
Resource management and cleanup for tool containers
Integration with existing Docker infrastructure
Use Case: Run custom tools in isolated Docker containers for enhanced security, enabling safe execution of untrusted code and ensuring consistent tool behavior across environments
2. Improve Consistency of Memory & Tool Reminders (@ncrispino)
Issue: #537
Enhance consistency of memory retrieval across agents
Improve tool reminder system for better agent awareness
Standardize memory access patterns
Optimize tool suggestion timing and relevance
Use Case: Ensure agents have consistent access to shared memory and receive relevant tool reminders at appropriate times, improving coordination and task execution quality
Success Criteria#
✅ Custom tools successfully launch in Docker containers
✅ Security isolation prevents tools from affecting host system
✅ Automatic cleanup of Docker resources after tool execution
✅ Tool execution performance comparable to native execution
✅ Memory retrieval is consistent across all agents
✅ Tool reminders appear at appropriate times with relevant suggestions
✅ Memory access patterns are standardized and documented
📋 v0.1.17 - Frontend Evaluation & MCP Ecosystem Expansion#
Features#
1. MassGen Terminal Evaluation (@ncrispino)
Issue: #476
Enable MassGen to evaluate and improve its own frontend/UI
Terminal session recording using asciinema for visual analysis
Automatic caption generation for recorded sessions
Video editing integration for demonstration materials
Comprehensive case study generation from terminal recordings
Self-improvement capabilities extended to frontend (currently backend-only via automation mode)
Use Case: Enable MassGen to analyze its own terminal interface, creating demonstration videos and documentation automatically, showcasing new features through automated workflows
2. Smithery MCP Tools Support (@ncrispino)
Issue: #521
Integration with Smithery to expand available MCP tools
Automatic discovery and installation of Smithery MCP servers
Curated registry of high-quality MCP tools from Smithery ecosystem
Simplified tool onboarding for users
Enhanced tool discovery and recommendation system
Use Case: Expand MassGen’s tool ecosystem by integrating with Smithery, giving users access to a wider range of curated MCP tools without manual configuration
Success Criteria#
✅ Terminal recording and playback system works reliably
✅ Video understanding capabilities accurately analyze terminal sessions
✅ Automated case study generation produces high-quality documentation
✅ MassGen successfully self-improves based on terminal analysis
✅ Smithery integration discovers and installs MCP tools automatically
✅ Tool registry includes curated Smithery tools with proper metadata
✅ Tool recommendation system suggests relevant Smithery tools
✅ Users can easily browse and install Smithery tools
🔨 Ongoing Work & Continuous Releases#
These features are being actively developed on separate parallel tracks and will ship incrementally on the MWF release schedule:
Track: Agent Adapter System (@Eric-Shang, ericshang.)#
PR: #283
Unified agent interface for easier backend integration
Shipping: Continuous improvements
Track: Irreversible Actions Safety (@franklinnwren, zhichengren)#
Human-in-the-loop approval system for dangerous operations
LLM-based tool risk detection
Target: v0.1.3 and beyond
Track: Multimodal Support (@qidanrui, danrui2020)#
PR: #252
Image, audio, video processing across backends
Shipping: Incremental improvements each release
Track: Memory Module (@qidanrui, @ncrispino, danrui2020, nickcrispino)#
Track: Agent Task Planning (@ncrispino, nickcrispino)#
Agent task planning with dependency tracking
Status: ✅ Completed in v0.1.7
Track: Automation & Meta-Coordination (@ncrispino, nickcrispino)#
LLM agent automation with status tracking and silent execution
MassGen running MassGen for self-improvement workflows
Status: ✅ Completed in v0.1.8
Case Study: Meta-level self-analysis demonstrating automation mode (
meta-self-analysis-automation-mode.md)
Track: DSPy Integration (@praneeth999, ram2561)#
Question paraphrasing for multi-agent diversity
Semantic validation and caching system
Status: ✅ Completed in v0.1.8
Track: Framework Streaming (@Eric-Shang, ericshang.)#
PR: #462
Real-time streaming for LangGraph and SmoLAgent intermediate steps
Enhanced debugging and monitoring for external framework tools
Status: ✅ Completed in v0.1.10
Track: Rate Limiting System (@AbhimanyuAryan, abhimanyuaryan)#
PR: #383
Multi-dimensional rate limiting for Gemini models
Model-specific limits with sliding window tracking
Status: ✅ Completed in v0.1.11
Track: MassGen Handbook (@a5507203, @Henry-811, crinvo, henry_weiqi)#
Issue: #387
Comprehensive user documentation and handbook at https://massgen.github.io/Handbook/
Centralized policies and resources for development and research teams
Status: ✅ Completed in v0.1.10
Track: Computer Use (@franklinnwren, zhichengren)#
PR: #402
Browser and desktop automation with OpenAI, Claude, and Gemini integration
Visual perception through screenshot processing and action execution
Status: ✅ Completed in v0.1.9
Track: Session Management (@ncrispino, nickcrispino)#
PR: #466
Complete session state tracking and restoration
Resume previous MassGen conversations with full context
Status: ✅ Completed in v0.1.9
Track: Semtools & Serena Skills (@ncrispino, nickcrispino)#
PR: #515
Semantic search capabilities via semtools (embedding-based similarity)
Symbol-level code understanding via serena (LSP integration)
Package as reusable skills within MassGen framework
Status: ✅ Completed in v0.1.12
Track: System Prompt Architecture (@ncrispino, nickcrispino)#
PR: #515
Complete refactoring of system prompt assembly
Hierarchical structure with improved LLM attention management
Skills system local execution support
Status: ✅ Completed in v0.1.12
Track: Multi-Agent Computer Use (@franklinnwren, zhichengren)#
PR: #513
Enhanced Gemini computer use with Docker integration
Multi-agent coordination for computer automation
VNC visualization and debugging support
Status: ✅ Completed in v0.1.12
Track: Code-Based Tools System / Automatic MCP Tool Selection (@ncrispino, nickcrispino)#
Issue: #414
Tool integration via importable Python code instead of schema-based tools
MCP server registry with auto-discovery
Reduces token usage through on-demand tool loading
Status: ✅ Completed in v0.1.13
Track: NLIP Integration (@praneeth999, @qidanrui, ram2561, danrui2020)#
PR: #475
Natural Language Integration Platform for advanced tool routing
Multi-backend support across Claude, Gemini, and OpenAI
Per-agent and orchestrator-level configuration
Status: ✅ Completed in v0.1.13
Track: Parallel Tool Execution (@praneeth999, ram2561)#
PR: #520
Configurable concurrent tool execution across all backends
Model-level and local execution controls
Asyncio-based scheduling with semaphore limits
Status: ✅ Completed in v0.1.14
Track: Gemini 3 Pro Support (@ncrispino, nickcrispino)#
PR: #530
Full integration for Google’s Gemini 3 Pro model
Function calling support with parallel tool capabilities
Status: ✅ Completed in v0.1.14
Track: Parallel File Operations (@ncrispino, nickcrispino)#
Issue: #441
Increase parallelism of file read operations
Standard efficiency evaluation and benchmarking methodology
Status: ✅ Completed in v0.1.14
Track: Launch Custom Tools in Docker (@ncrispino, nickcrispino)#
Issue: #510
Enable custom tools to run in isolated Docker containers
Security isolation and portability for custom tool execution
Target: v0.1.16
Track: Improve Consistency of Memory & Tool Reminders (@ncrispino, nickcrispino)#
Issue: #537
Enhance consistency of memory retrieval across agents
Improve tool reminder system for better agent awareness
Standardize memory access patterns
Target: v0.1.16
Track: RL Integration (@qidanrui, @praneeth999, danrui2020, ram2561)#
Issue: #527
Reinforcement learning integration for agent optimization
Adaptive agent behavior based on feedback and outcomes
Reward modeling for multi-agent coordination
Target: v0.1.15
Track: Git Worktrees for Multi-Agent (@ncrispino, nickcrispino)#
Issue: #514
Enable multiple agents to work on different Git worktrees simultaneously
Isolated working directories for parallel agent development
Target: v0.1.15
Track: MassGen Terminal Evaluation (@ncrispino, nickcrispino)#
Issue: #476
Self-evaluation and improvement of frontend/UI through terminal recording
Automated video generation and case study creation
Target: v0.1.17
Track: Smithery MCP Tools Support (@ncrispino, nickcrispino)#
Issue: #521
Integration with Smithery to expand available MCP tools
Automatic discovery and installation of Smithery MCP servers
Target: v0.1.17
Track: Coding Agent Enhancements (@ncrispino, nickcrispino)#
PR: #251
Enhanced file operations and workspace management
Shipping: Continuous improvement
Track: Web UI (@voidcenter, justin_zhang)#
PR: #257
Visual multi-agent coordination interface
Target: TBD
🎯 Long-Term Vision (v0.2.0+)#
Advanced Orchestration Patterns
Task/subtask decomposition and parallel coordination
Assignment of agents to specific tasks and increasing of diversity
Improvement in voting as tasks continue
Visual Workflow Designer
No-code multi-agent workflow creation
Drag-and-drop agent configuration
Real-time testing and debugging
Enterprise Features
Role-based access control (RBAC)
Audit logs and compliance reporting
Multi-user collaboration
Advanced analytics and cost tracking
Additional Framework Integrations
LangChain agent support
CrewAI compatibility
Custom framework adapters
Complete Multimodal Pipeline
End-to-end audio processing (speech-to-text, text-to-speech)
Video understanding and generation
Advanced document processing (PDF, Word, Excel)
🔗 GitHub Integration#
Track development progress:
🤝 Contributing#
Interested in contributing? You have two options:
Option 1: Join an Existing Track
See Contributors & Contact table above for active tracks
Contact the track owner via Discord to discuss your ideas
Follow CONTRIBUTING.md for development process
Option 2: Create Your Own Track
Have a significant feature idea? Propose a new track!
Reach out via the #massgen channel on Discord
Work with the MassGen dev team to integrate your track into the roadmap
Become a track owner and guide other contributors
See CONTRIBUTING.md for development setup, code standards, testing, and documentation requirements.
See Also#
Contributing to MassGen - Contribution guidelines with documentation requirements
Changelog - Complete release history
—