AVM

1/ Most AI agents in crypto can talk. But they can’t actually operate. They live inside prompts instead of real environments where they can run code, access wallets, and interact with the ecosystem. That’s the gap we built $AVM to solve.

Agents shouldn’t live only in prompts With AVM they run inside isolated sandboxes, with terminal access, file systems, and wallet connectivity.

AVM is the infrastructure for crypto-native agents. Spin up sandboxes, attach wallets, run real code and interact with onchain systems from actual execution environments. App.avm.codes

7/ If you're building agents, automation or onchain tools, AVM gives you the missing execution layer: ☄️run code ☄️manage wallets ☄️automate onchain actions ☄️operate outside the prompt Build agents that actually do things 👉app.avm.codes

6/ What’s coming next: ➡️Agents operating directly on the crypto platforms people already use ➡️More persistent environments for long-running agents ➡️x402-native flows where agents can pay, get paid, and access services autonomously The goal is simple: make web3 agents actually usable.

1/ Most AI agents in crypto can talk. But they can’t actually operate. They live inside prompts instead of real environments where they can run code, access wallets, and interact with the ecosystem. That’s the gap we built $AVM to solve.

Run Claude Code in the terminal with your own API. Spin up multiple Claude Code instances. Orchestrate them across isolated sandboxes. This is the future AVM is building.

The vote on the DAO proposal has concluded with a YES. Thank you to everyone who voted and exercised their rights. snapshot.org/#/s:avmdao.eth…

Last day to vote on the AVM DAO proposal.

DAO proposal: Starting March 1, 2026, AVM platform inflows will be used for AVM token buybacks, subject to DAO approval. This directly aligns platform usage with token value support. Vote before Feb 28, 2026 · 12:20 PM ✅ Recommendation: YES

The first AVM DAO proposal is out. Read and vote to help us shape the future of AVM! snapshot.org/#/s:avmdao.eth…

The AVM roadmap is live. Updated and aligned with our long-term vision roadmap.avm.codes

📌AVM V2 is live in production! This release is a major upgrade to how AVM handles agent execution, sandboxing, and developer workflows. V2 introduces a more opinionated, production-grade execution layer built for real-world agent workloads, with stronger guarantees around isolation, control, and usability. On the infrastructure side, sandbox enforcement is now deterministic across CPU, memory, and instance limits, ensuring predictable behavior, especially for the free tier. The runtime filesystem now supports full bidirectional file transfers, allowing artifacts, datasets, and outputs to persist throughout a sandbox lifecycle. Sandbox provisioning includes deeper pre-runtime validation, and environment variables are now managed through a structured, deployment-style model with protected variables in Pro mode and explicit key–value control. From a UX perspective, developers can interact with live sandboxes (up to 5 for free) via an embedded terminal, explore the filesystem natively, and move files in and out directly from the UI. $AVM V2 brings us closer to our goal: a reliable, developer-first execution layer for autonomous agents. Live now: platform.avm.codes 🛠️Next steps: - Gather developer feedback and onboard users - Expand the team with industry experts to accelerate growth in Web3

The AVM platform is live and we’re building it every week. Long road, long vision. 2026 is the year of sandboxes!

2026 is the year of sandboxes. $AVM is the execution layer for AI agents, no infrastructure to manage. Agents install dependencies, run in isolated environments, and terminate automatically. Just verifiable execution.

Spot On @Rasmic