Roadmap
This roadmap outlines planned features and improvements for both the mcp-agent framework and MCP Agent Cloud platform. We aim to build the most comprehensive platform for developing and deploying AI agents using the Model Context Protocol.Framework Roadmap
v1.0.0 Release (September 2025)
The first stable release will establish mcp-agent as the standard framework for building MCP-based agents.Core Features
- TypeScript SDK - Full TypeScript/JavaScript support for agent development
- Advanced Tool Selection - Intelligent filtering for MCP tools with large payloads
- Smart tool relevance scoring
- Context-aware tool filtering
- Middleware for processing tool results before LLM consumption
- Full MCP Support - Complete implementation of MCP specification
- OAuth 2.1 authentication support
- URL elicitation for dynamic resource discovery
- Sampling API for server-side LLM invocation
- Multimodality - Native support for multimodal interactions
- Image processing and generation
- Audio/speech capabilities
- Document understanding
- Streaming & Events - Real-time capabilities
- Server-sent events for long-running operations
- WebSocket subscriptions for state changes
- Event-driven architecture for reactive agents
Memory Management
- Procedural Memory - Agents learn and remember problem-solving patterns
- Task completion strategies
- Error recovery procedures
- Performance optimization patterns
- Session Memory - Conversation and context persistence
- Cross-session continuity
- Context summarization
- Relevance-based memory retrieval
- User Memory - Personalized agent experiences
- User preference learning
- Interaction history
- Adaptive behavior based on user patterns
- Tree-based Memory Scoping - Hierarchical memory organization
- Memory inheritance between agent scopes
- Selective memory sharing
- Privacy-preserving memory boundaries
Agent Identity Management
- IAM-style Scopes - Hierarchical agent permissions
- Role-based access control
- Capability inheritance
- Fine-grained permission management
- Agent Identities - Distinct agent personas and capabilities
- Identity switching for multi-role agents
- Capability composition
- Trust boundaries between agents
Human-in-the-Loop
- Cleaner Semantics - Improved APIs for human interaction
- Declarative approval workflows
- Structured feedback collection
- Progressive disclosure of complexity
- Built-in UI Components - Ready-to-use interaction patterns
- Approval dialogs
- Feedback forms
- Progress indicators
Beyond v1.0.0
Self-Improving Agents
- Reinforcement Learning - Agents that optimize through experience
- A/B Testing Framework - Systematic improvement through experimentation
- Performance Analytics - Detailed metrics for agent optimization
Advanced Workflow Patterns
- Distributed Workflows - Multi-region workflow execution
- Workflow Composition - Building complex workflows from simpler ones
- Dynamic Workflow Generation - LLM-generated workflow definitions
MCP Agent Cloud Roadmap
Near-term (Q1 2025)
OAuth 2.1 Authorization Server
- Managed Auth Service - Turnkey OAuth 2.1 implementation
- Support for major identity providers (Google, GitHub, Microsoft)
- Custom identity provider integration
- Fine-grained access control
- Token management and rotation
- MCP Auth Specification - Full compliance with MCP authorization spec
- Dynamic client registration
- Resource metadata endpoints
- Bearer token validation
Transport Enhancements
- Streamable HTTP Support - Currently SSE-only
- Full HTTP/2 streaming
- Chunked transfer encoding
- Response streaming for large payloads
Agent Registry
- Public Agent Marketplace - Discover and deploy community agents
- Verified publisher program
- Usage analytics and ratings
- One-click deployment
- Revenue sharing for premium agents
- Private Registries - Enterprise agent catalogs
- Internal agent sharing
- Compliance and governance
- Version control and rollback
Mid-term (Q2-Q3 2025)
Enterprise Features
- Single Sign-On (SSO) - Enterprise authentication
- SAML 2.0 support
- OpenID Connect
- Active Directory integration
- Audit & Compliance - Enterprise-grade governance
- Comprehensive audit logs
- Compliance reporting (SOC2, HIPAA)
- Data residency controls
- Private Cloud Deployment - On-premises options
- Kubernetes operators
- Terraform modules
- Air-gapped deployments
Advanced Orchestration
- Workflow Templates - Reusable workflow patterns
- Scheduled Workflows - Cron-based execution
- Event-Driven Triggers - Webhook and event-based activation
- Workflow Versioning - Blue-green deployments for workflows
Observability & Monitoring
- Advanced Tracing - Distributed tracing across agents
- Custom Metrics - Application-specific monitoring
- Alerting Rules - Proactive issue detection
- Performance Profiling - Bottleneck identification
Long-term (Q4 2025+)
Global Infrastructure
- Multi-Region Deployment - Global agent distribution
- Automatic failover
- Geo-routing
- Data sovereignty compliance
- Edge Deployment - Low-latency agent execution
- CDN integration
- Regional caching
- Offline-capable agents
Developer Experience
- Visual Workflow Designer - No-code agent creation
- Agent Testing Framework - Comprehensive testing tools
- Debugging Tools - Step-through debugging for workflows
- Performance Optimization - Automatic agent optimization
Platform Ecosystem
- Plugin System - Extensible platform capabilities
- Partner Integrations - Deep integration with cloud providers
- SDK for Multiple Languages - Beyond Python and TypeScript
- GraphQL API - Flexible data access
Community & Ecosystem
Documentation & Learning
- Interactive Tutorials - Hands-on learning experiences
- Video Course Series - Comprehensive agent development training
- Certification Program - Professional certification for developers
- Best Practices Guide - Production-ready patterns and anti-patterns
Developer Tools
- VS Code Extension - Enhanced development experience
- CLI Enhancements - More powerful command-line tools
- GitHub Actions - CI/CD integration
- Testing Frameworks - Unit and integration testing tools
Community Building
- Developer Forum - Community support and discussion
- Monthly Webinars - Feature updates and tutorials
- Hackathons - Community challenges and prizes
- Open Source Contributions - Community-driven development
Feedback & Contribution
We welcome community feedback and contributions to shape this roadmap:- GitHub Issues: Report bugs and request features
- Discord Community: Join discussions
- Email: team@lastmileai.dev
Version History
| Version | Release Date | Highlights |
|---|---|---|
| v0.1.0 | December 2024 | Initial alpha release |
| v0.2.0 | January 2025 | Temporal integration |
| v0.3.0 | February 2025 | Cloud deployment beta |
| v0.4.0 | March 2025 | TypeScript SDK alpha |
| v0.5.0 | April 2025 | Memory management preview |
| v1.0.0 | September 2025 | First stable release |
Commitment to Standards
mcp-agent is committed to maintaining full compatibility with the Model Context Protocol specification. As MCP evolves, we will:- Track and implement new MCP features promptly
- Contribute to MCP specification development
- Maintain backward compatibility
- Provide migration guides for breaking changes