Recently, OpenClaw and PicoClaw frameworks have been successfully deployed on the VisionFive 2 single board computer(SBC), granting AI agents an “RISC-V body.” On this basis, leveraging the underlying key management and payment capabilities of the BoAT SDK, the VisionFive 2 has been upgraded into an intelligent economic entity capable of independent signing, autonomous reasoning, on-chain attestation, and USDC settlement. A full-stack closed loop of “Self-Owned Machine Economy (SOME)”, built on open-source hardware, has now moved from theory to practice.
Deploying OpenClaw: Injecting an “AI Brain” into RISC-V Hardware
The first step toward machine autonomy is endowing it with the ability to understand and execute natural language tasks. On the VisionFive 2 running Debian, developers can fully deploy the OpenClaw agent framework.
Since the official Node.js RISC-V prebuilt package is not available, the process began with compiling and installing Node.js natively on the riscv64 architecture from source code. Afterwards, by configuring API keys for cloud-based LLMs such as Alibaba Bailian (e.g., Qwen Plus), the VisionFive 2 successfully connects to powerful cloud LLMs.
Once deployed, users can interact directly with an Agent on the VisionFive 2 via command-line terminal, or through the OpenClaw Gateway Dashboard for visual UI interaction, giving this RISC-V SBC basic automated task processing capabilities.
Chatting with OpenClaw Dashboard in the UI:
The tutorial has been published on the RVspace forum, detailing the entire process of deploying OpenClaw on the VisionFive 2 to help users quickly achieve AI conversation and automated task handling. The complete official technical documentation is being prepared and will be available soon.
Integrating PicoClaw and Telegram for Edge Remote Intelligent Control
With fully-functional OpenClaw in place, to accommodate more flexible edge-side interactions, the project further introduces the ultra-lightweight AI assistant PicoClaw and binds it to the widely-used social platform Telegram. Two VisionFive 2 devices, each holding its own independent key, completed a real transaction in which one sold sensor data and the other purchased it using USDC. From underlying hardware to top-level payment, the entire process required no human intervention – the devices executed precise micropayments for the services they consumed.
By binding PicoClaw to Telegram, operational workflows have been dramatically simplified. Users simply open Telegram and send a brief chat message (e.g., “Please run boat-attest-c to attest memory/CPU usage”). Upon parsing the user’s Telegram command, PicoClaw triggers the VisionFive 2 to invoke the boat-attest-c program, which is natively compiled in pure C.
Sending a Telegram message to PicoClaw on VisionFive 2 to trigger the attestation workflow:
Integrating BoAT SDK: Closing the Loop on Machine Autonomous Payments
The BoAT (Blockchain of AI Things) SDK directly “implants” blockchain capabilities into AI agents. As a non-custodial underlying library, it enables devices to perform local key generation and x402 transaction signing without relying on the cloud. Under the Linux environment of VisionFive 2, BoAT fully supports the core steps from identity definition to autonomous payment:
(1) Data Collection and Local Signing: The VisionFive 2 autonomously collects current system core metrics (such as memory usage and CPU load), and performs a hash signature using the Ed25519 identity key generated locally on the device.
(2) On-Chain Tamper-Proof Attestation: The signed data is securely submitted to the HashAnchor platform and anchored to blockchains such as Polygon, forming an immutable proof.
(3) x402 Protocol Autonomous Payment: When the attestation service returns an HTTP 402 Payment Required response, the magic happens – the device independently signs the authorization using its local Secp256k1 wallet key. Subsequently, two physical devices complete a real USDC stablecoin payment settlement via standard HTTP protocol through the Circle Gateway. The entire process is fully automated with no human intervention, enabling devices to execute precise micropayments down to the cent for the services they use.
Verifying attestation hashes on HashAnchor:
This achievement validates the feasibility of the “SOME Five-Layer Open Architecture”: from open-source hardware (VisionFive 2), open-source agents (Claw), open-source wallets (BoAT MWR), open-source data (HashAnchor), to open-source currency (USDC and x402). It proves that on RISC-V open hardware, a complete cryptographic and blockchain technology stack can run natively without relying on cloud custody or any proprietary lock-in.
To let more enthusiasts experience this cutting edge capability, the complete source code, build instructions, and deployment scripts are available on GitHub. You can complete the above deployment on your own VisionFive 2 and unlock the capability of autonomous machine payments.