Roadmap
AVM Strategic Roadmap
Overview: This document details the strategic roadmap for the AVM protocol, outlining the path from MVP to global agent-native marketplace.
Abstract
This roadmap defines a five-phase plan for the AVM protocol, reflecting both present status and long-term strategic vision. Phase 1 covers foundational deliverables—MVP dashboard and beta SDK release. Subsequent phases focus on token launch, decentralization, scalable global adoption, and the final stage of agent-native execution infrastructure.
1. Introduction
AVM advances through a structured timeline to establish robust compute capabilities, expand functionality, decentralize network operations, and scale global adoption. The protocol’s MVP dashboard is imminent and the SDK is live for beta testers; the token launch is scheduled for Phase 2. The final phase outlines the long-term goal of cementing AVM as the global compute backbone for autonomous agents.
2. Development Phases
| Phase | Name | Timeline | Goal |
|---|---|---|---|
| 1 | Foundation | Q3 2025 | MVP dashboard + SDK |
| 2 | Expansion | Q4 2025 | 1,000 developers |
| 3 | Decentralization | Q1-Q2 2026 | Permissionless node network, 100 nodes |
| 4 | Scaling | Q3-Q4 2026 | 10,000 users, ecosystem integration |
| 5 | Ascension | 2027 – Q3 2028 | Global agent-native marketplace |
🚀 Phase 1: Foundation (Q3 2025)
✅ Completed:
- Python virtual machine deployment
- TypeScript virtual machine deployment
- Tools reuse infrastructure
- Reusable sandboxes (permanent tools)
- Core execution capabilities and SDK for beta testers
- Initial web dashboard for monitoring compute tasks
- MCP integrations with leading AI automation platforms
🔄 In Progress:
- Security, monitoring, and developer support infrastructure
- PHP virtual machine deployment
- Go virtual machine deployment
🌍 Phase 2: Expansion (Q4 2025)
- Sandboxes Marketplace: Developers can now monetize their custom toolkits
- Real-time analytics and reporting dashboards
- Ecosystem grants and strategic partnerships
- Multi-modality support: Enable output of images and videos, not just text
Phase 3: Decentralization (Q1-Q2 2026)
- Self-hosted virtual machines: Enable users to run their own AVM nodes
- AVM consensus mechanism: Coordinate decentralized compute providers
- Transition to fully permissionless, decentralized execution network
- Launch DAO governance modules for protocol parameter management
- Tokenized incentives for compute providers
Phase 4: Scaling (Q3-Q4 2026)
- Distributed storage: Enable sandboxes to save and reuse files for long-term memory
- Execute full-scale marketing and adoption campaigns
- Expand global node infrastructure for AI agent networks
- Advanced performance optimization (GPU support, dynamic load balancing)
Phase 5: Ascension (2027 – Q3 2028)
Cement AVM as the global backbone for AI execution.
- Fully equipped virtual machines: Complete desktop experience where agents have their own personal computer and achieve real-world tasks with minimal assistance
- AVM Code: Claude Code-style development environment enabling agentic coding loops from zero to goal
- Mature agent-native marketplace for compute, tools, and automation
3. Strategic Development Plan
The initial AVM deployment (v0) is hosted on distributed cloud infrastructure, with an alpha dashboard for task monitoring. Future versions will incorporate a proprietary VM Solver, enabling peer-to-peer contributions of compute resources for LLM processing, enhancing network resilience and reducing centralization.
4. Developer Integration
Developers integrate AVM via:
- High-Performance HTTP API — Direct LLM-to-AVM solver communication for efficient task execution.
- Comprehensive SDKs — Available for TypeScript (
@avm-ai/avm-vercel-ai) and MCP server (@avm-ai/avm-mcp), supporting Python code execution through the Vercel AI SDK and MCP-compatible clients.