DefenseClaw Edge Connector

Cisco's open-source DefenseClaw gains an Edge Connector that allows or blocks AI agent tool calls directly on IoT and robot hardware.

Cisco · Security and safety tooling · security

Published in Feed

Overview

On October 14, 2026, Cisco described an Edge Connector for its open-source DefenseClaw project, extending AI agent security governance to physical devices. The component runs on any Linux device, including Raspberry Pi and NVIDIA Jetson, intercepts each tool call an agent makes, and returns ALLOW, BLOCK, or ESCALATE in microseconds. Unknown tools escalate to the main DefenseClaw instance, and if that instance is unreachable the action is denied. Every decision is written to a tamper-evident audit trail on the device.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Intercepts agent tool calls on-device and returns ALLOW, BLOCK, or ESCALATE
  • Blocks denied destinations, known-bad tool hashes, and prompt-injection patterns locally
  • Prompt injection detected before the LLM processes input, so no inference tokens are spent
  • Escalates unknown tools to the main DefenseClaw instance; denies the action if it is unreachable
  • Tamper-evident on-device audit log
  • Runs on any Linux device, including Raspberry Pi and NVIDIA Jetson

Use cases

  • Governing tool use by robots and IoT agents at the edge
  • Enforcing local policy on agents with intermittent connectivity
  • Auditable agent actions on embedded hardware

Pricing

Ships as part of the DefenseClaw open-source project; no pricing published.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
14 Oct 2026
Record updated
16 Oct 2026
Record last checked
16 Oct 2026

Explore the Feed

Other entries filed under security and safety tooling; inclusion does not imply equivalent features.

  • Lakera Guard 2.0

    Lakera's real-time guardrail API for LLM apps, updated with multi-agent protection and a new benchmark for detecting indirect prompt injection in tool outputs.

  • Fleming-1

    Sierra's model that listens to a live phone call and scores whether the caller is an AI agent rather than a human.

  • Anthropic Cyber Mission

    Anthropic's long-term program giving Claude models, engineers, and threat research to defenders of critical infrastructure and open-source software.

Browse the Feed