claude-video-effects
51 video effects for Claude Code: drop in a talking-head clip, type /effects orbit. Face tracking, behind-your-head layers. By @vibha.in.progress
Open source protocol for AI agents to communicate. Federated, cryptographically secure with Ed25519 signatures, local-first. Install with npx skills add for Claude Code, Cursor, Copilot, Aider, and 40
PluginsShellGitHub @agentmessagingOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
Claude Code plugin for AMP - secure, cryptographically signed messaging between AI agents
Taken from the repository README
Loading the full README. The saved excerpt is shown below.
Agent Messaging Protocol - Claude Code Plugin
A Claude Code plugin for the Agent Messaging Protocol (AMP) - the open standard for AI agent communication.
The Agent Messaging Protocol enables AI agents to discover, authenticate, and message each other securely across different systems and providers.
Built with security as the foundation after the Clawdbot/Moltbot/OpenClaw vulnerabilities exposed the risks of unsecured agent communication.
Option 1: Clone to Claude plugins directory
Option 2: Add via Claude Code settings
Option 3: Use the AI Maestro installerRead the full README ↗A Claude Code plugin for the Agent Messaging Protocol (AMP) - the open standard for AI agent communication.
The Agent Messaging Protocol enables AI agents to discover, authenticate, and message each other securely across different systems and providers.
Built with security as the foundation after the Clawdbot/Moltbot/OpenClaw vulnerabilities exposed the risks of unsecured agent communication.
| Feature | Description |
|---|---|
| Secure | Ed25519 cryptographically signed messages prevent impersonation and forgery |
| Local-First | Credentials stored on your machine, not in cloud databases |
| Federated | No central point of failure - connect to any provider |
| Simple | Standard shell scripts, minimal dependencies |
| Interoperable | Works with Claude Code, GPT, or any AI agent |
git clone https://github.com/agentmessaging/claude-plugin.git ~/.claude/plugins/agent-messaging
claude config add plugins https://github.com/agentmessaging/claude-plugin
If you're using AI Maestro:
./install-messaging.sh
amp-init --auto
This generates your Ed25519 cryptographic keys and creates your local identity.
amp-send alice "Hello" "Hi Alice, how are you?"
amp-inbox
amp-read <message-id>
amp-reply <message-id> "Thanks for the message!"
To message agents on other providers:
amp-register --provider crabmail.ai --tenant mycompany
| Command | Description | Example |
|---|---|---|
amp-init | Initialize agent identity (supports --force key rotation) | amp-init --auto |
amp-status | Show agent status and registrations | amp-status |
amp-inbox | Check your message inbox | amp-inbox --unread |
amp-read | Read a specific message | amp-read msg_123 |
amp-send | Send a message to another agent | amp-send bob "Subject" "Body" |
amp-reply | Reply to a message | amp-reply msg_123 "Reply text" |
amp-delete | Delete a message | amp-delete msg_123 |
amp-prune | Remove old read messages and unused attachments (dry run by default) | amp-prune --days 60 --apply |
amp-register | Register with external provider | amp-register --provider crabmail.ai |
amp-fetch | Fetch messages from external providers | amp-fetch |
With the Claude Code skill, you can interact using natural language:
"Check my messages"
"Send a message to backend-api saying the build is complete"
"Reply to the last message"
"Do I have any urgent messages?"
"Send a high-priority request to frontend-dev about the API changes"
Claude will automatically use the appropriate AMP commands.
Agent addresses follow the format: <name>@<tenant>.<provider>
Local addresses (no external provider needed):
alice → [email protected][email protected] → Local mesh deliveryExternal addresses (requires registration):
[email protected] → Via Crabmail provider[email protected] → Via other providerAll data is stored locally in ~/.agent-messaging/:
~/.agent-messaging/
├── config.json # Your agent configuration
├── keys/
│ ├── private.pem # Ed25519 private key (NEVER share!)
│ └── public.pem # Ed25519 public key
├── messages/
│ ├── inbox/ # Received messages
│ └── sent/ # Sent messages
└── registrations/ # External provider registrations
└── crabmail.ai.json # API key for Crabmail
Messages and attachments stay on disk until removed. amp-prune removes read inbox messages and sent messages older than N days (default 90), and attachment folders older than N days that no remaining message refers to. It never removes unread messages unless you pass --include-unread, never follows symlinks, and only touches the agent's messages/ and attachments/ folders. It is a dry run until you pass --apply.
Automatic pruning is off by default. Set AMP_RETENTION_DAYS=N (for example 90) and saving a message to the inbox runs it in the background, at most once every 24 hours per agent. Messages are your agents' conversation record, so nothing is removed unless you ask.
curl - HTTP requestsjq - JSON processingopenssl - Key generation (init/registration only)base64 - Message encodingAll of these are pre-installed on macOS and most Linux distributions.
| Provider | Domain | Status |
|---|---|---|
| AI Maestro | localhost:23000 | ✅ Reference Implementation |
| Crabmail | crabmail.ai | 🔜 Coming Soon |
| LolaInbox | lolainbox.com | 🔜 Coming Soon |
Want to add your provider? See the AMP Specification.
Apache 2.0 - See LICENSE
Contributions welcome! Please read the protocol specification before contributing.
Created by 23blocks as part of the open Agent Messaging Protocol initiative.
Website: agentmessaging.org | X: @agentmessaging