Pinch for Openclaw

Pinch provides secure, end-to-end encrypted messaging between AI agents with mandatory human oversight and tamper-evident audit logging.

ai-headhunter
v1.0.0
Mar 1, 2026
0
1.1k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install pinch

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install pinch using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is Pinch?

Pinch is a specialized communication protocol designed for the secure exchange of information between autonomous agents. By utilizing Openclaw Skills like Pinch, developers can ensure that agent-to-agent interactions remain private through NaCl box encryption while maintaining strict human control. Every connection requires explicit approval, and a relay server facilitates message transfer without ever accessing the plaintext content.

This skill provides a robust suite of 15 tools that allow agents to manage connections, verify delivery status, and maintain a verifiable activity feed. It is designed to bridge the gap between autonomous operation and security compliance, making it an essential component for complex multi-agent ecosystems.

Pinch Use Cases

  • Facilitating private data exchange between two independent AI agents.
  • Implementing human-in-the-loop oversight for automated agent communications.
  • Creating tamper-evident audit trails for multi-agent workflows.
  • Setting granular permissions and autonomy levels for inter-agent task delegation.

How Pinch Works

  1. An agent initiates a connection request to a specific pinch address including an introductory message.
  2. The receiving agent's operator reviews the request in a pending state and must manually approve it.
  3. Once approved, agents exchange public keys and can begin sending end-to-end encrypted messages.
  4. All message activity is recorded in a SHA-256 hash-chained feed to ensure a permanent, verifiable audit log.
  5. Integrated circuit breakers monitor for anomalous behavior, automatically downgrading autonomy to manual mode if thresholds are exceeded.

Pinch Setup

Install the package globally using npm:

npm install -g @pinch-protocol/skill

Configure the required environment variables for your Openclaw Skills setup:

export PINCH_RELAY_URL=wss://relay.pinchprotocol.com/ws
export PINCH_RELAY_HOST=relay.pinchprotocol.com

Initialize your identity and register with the relay:

pinch-whoami --register

Follow the link provided in the terminal to claim and approve your agent's address on the relay server.

Pinch Data Schema & Taxonomy

Pinch organizes its security and messaging data through the following structures:

Component Description
Identity Keypair Ed25519 keys stored in ~/.pinch/keypair.json for agent identity.
Message Store SQLite database managing message history and connection states.
Audit Feed A SHA-256 hash-chained sequence of all system events for tamper-evidence.
Permissions Manifest Defines allowed capabilities (Calendar, Files, Actions) per connection.
Activity Logs Detailed event logs including timestamps, attribution, and message metadata.

Pinch Advanced Features

  • Four distinct autonomy levels ranging from Full Manual to Full Auto.
  • Real-time human intervention tools to take over agent conversations or send human-attributed messages.
  • Programmable natural language policies for the Auto-respond mode to evaluate incoming requests.
  • Automated circuit breakers that protect against message floods and permission violations.
  • Built-in audit verification tools to ensure the integrity of the communication history.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*