1Claw launches zero‑trust AI agent security platform with Bankr partnership
*1Claw launched a zero‑trust security platform for AI agents, adding integration with Bankr and major cloud partners.*
Also in: 日本語 · エージェント経済 · 한국어 · 에이전트포스트 · 简体中文 · 链智日报
1Claw announced on October 6, 2026 a zero-trust framework that prevents agents from accessing sensitive data while executing transactions and API calls. The platform encrypts credentials in hardware and processes operations server-side, eliminating exposure in agent prompts or logs.
The 1Claw Platform API includes three components, according to the company’s documentation. Shroud scans for malicious code in agent traffic. The Intents API signs transactions and executes API calls without exposing keys. Automations and Runtimes provide secure, autonomous execution environments. The platform supports 109 Ethereum Virtual Machine (EVM) chains and additional blockchains such as Bitcoin, Solana, XRP, Cardano, and Tron.
The system assigns each agent an identity and scoped permissions, with keys stored in hardware security modules (HSMs) and traffic inspected in trusted execution environments (TEEs). Secrets are encrypted with per-secret data encryption keys wrapped by cloud key management services, optionally split across Google Cloud Platform, AWS, and Azure. The company says this prevents prompt injection attacks, API key leaks, and exfiltration attempts.
The platform integrates with development tools including Claude, Cursor, and GPT, replacing environment files with runtime credential fetching. For on-chain operations, DeFi agents sign transactions through the Intents API, with keys remaining HSM-wrapped or stored in user passkeys. Customer-facing agents receive scoped credentials, and every action generates an audit log entry.
1Claw disclosed a partnership with Bankr and listed partners Google Cloud, Circle, Tenderly, Coder, and the XRP Ledger Foundation. The company offers a free tier for new users.
According to 1Claw’s site, the platform provides bi-directional injection scoring, exfiltration URL blocking, and multi-layer threat detection covering encoding, network, filesystem, and social engineering vectors. Agents can be configured with token caps, blocked domains, and spend limits. The Intents API supports allowlists, daily transaction limits, and Tenderly simulation for pre-execution validation.
The company’s documentation states that secrets are never exposed in agent context windows, logs, or memory. Instead, agents execute through server-side bindings, with credentials injected at runtime. The platform also offers embedded wallets on six blockchains, spend policies, and OAuth-based authentication.