AgentAnycast

v0.3.2Communicationstable

Discover and communicate with AI agents over encrypted P2P networks. Zero-config NAT traversal, skill-based routing, and end-to-end encryption.

a2aa2a-protocolagent-communicationagent-to-agentai-agent
Share:
76
Stars
0
Downloads
0
Weekly
0/5

What is AgentAnycast?

AgentAnycast is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to discover and communicate with ai agents over encrypted p2p networks. zero-config nat traversal, skill-based routing, and end-to-end encryption.

Discover and communicate with AI agents over encrypted P2P networks. Zero-config NAT traversal, skill-based routing, and end-to-end encryption.

This server falls under the Communication category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • Discover and communicate with AI agents over encrypted P2P n

Use Cases

Discover and communicate with AI agents over encrypted P2P networks.
Enable zero-config NAT traversal with skill-based routing.
AgentAnycast

Maintainer

LicenseApache-2.0
Languageshell
Versionv0.3.2
UpdatedMay 9, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

NPM

npx -y agentanycast

Manual Installation

npx -y agentanycast

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use AgentAnycast

AgentAnycast is a peer-to-peer agent communication network that lets AI agents discover and talk to each other over encrypted connections without manual network configuration. It provides zero-config NAT traversal using mDNS for local networks and optional relay nodes for cross-network routing, combined with NaCl box end-to-end encryption and W3C DID-based identity. As an MCP server it exposes tools for discovering agents by skill, routing tasks to peers, and monitoring execution status — making it a building block for multi-agent architectures where AI assistants need to delegate work to specialized agents.

Prerequisites

  • Node.js (for the npx CLI) or Go toolchain (to build from source)
  • AgentAnycast daemon binary or npm package installed
  • An MCP-compatible client such as Claude Desktop, Cursor, or VS Code
  • Optional: a relay node address for cross-network (non-local) agent discovery
1

Install the AgentAnycast daemon

Install the npm package globally or use npx to run it directly. Alternatively, build from source using the Go toolchain if you need the full binary.

npm install -g agentanycast
# or build from source:
git clone https://github.com/AgentAnycast/agentanycast-node
cd agentanycast-node
go build -o agentanycastd ./cmd/agentanycastd
2

Start the daemon in MCP mode

Launch the daemon with the -mcp flag to expose it as an MCP server over stdio. On a local network, agents will auto-discover each other via mDNS with no additional configuration.

agentanycastd -mcp
3

Configure your MCP client

Add AgentAnycast to your MCP client configuration file. The daemon communicates over stdio and stores keys and data in ~/.agentanycast/ by default.

{
  "mcpServers": {
    "agentanycast": {
      "command": "npx",
      "args": ["-y", "agentanycast", "-mcp"]
    }
  }
}
4

(Optional) Connect to cross-network agents via relay

To reach agents outside your local network, start the daemon with a bootstrap-peers relay address. The relay handles NAT traversal so agents on different networks can discover and communicate with each other.

agentanycastd -mcp -bootstrap-peers "/ip4/<RELAY_IP>/tcp/4001/p2p/12D3KooW..."
5

(Optional) Wrap an existing MCP server as an agent

Use the --mcp-proxy flag to expose any existing MCP server (such as a filesystem server) as a discoverable AgentAnycast agent that remote peers can call by skill.

agentanycastd --mcp-proxy "npx -y @modelcontextprotocol/server-filesystem /home/user"

AgentAnycast Examples

Client configuration

Claude Desktop configuration for AgentAnycast in MCP stdio mode with optional environment variable overrides.

{
  "mcpServers": {
    "agentanycast": {
      "command": "npx",
      "args": ["-y", "agentanycast", "-mcp"],
      "env": {
        "AGENTANYCAST_LOG_LEVEL": "info",
        "AGENTANYCAST_ENABLE_MDNS": "true"
      }
    }
  }
}

Prompts to try

Example prompts for discovering agents and delegating tasks through the AgentAnycast network.

- "Discover all agents on the local network and list their skills."
- "Find an agent that can handle code-review tasks and route this pull request to it."
- "What is the status of the task I sent to the data-analysis agent?"
- "List all peers currently reachable on the AgentAnycast network."
- "Query the agent card for peer 12D3KooW... and show me its capabilities."

Troubleshooting AgentAnycast

No agents discovered on the local network

Ensure AGENTANYCAST_ENABLE_MDNS is set to true (it is by default). Check that all agents are on the same subnet and that mDNS traffic (UDP port 5353) is not blocked by a firewall. Verify the daemon is running with agentanycastd status.

Cross-network agents are unreachable

Provide a bootstrap-peers relay address using the -bootstrap-peers flag or the bootstrap_peers key in ~/.agentanycast/config.toml. Deploy your own relay or use a community relay endpoint.

MCP client cannot connect to the daemon

Confirm the daemon is running in MCP mode (-mcp flag) and that AGENTANYCAST_MCP_LISTEN is not set to a conflicting address. Check AGENTANYCAST_LOG_LEVEL=debug for detailed connection logs.

Frequently Asked Questions about AgentAnycast

What is AgentAnycast?

AgentAnycast is a Model Context Protocol (MCP) server that discover and communicate with ai agents over encrypted p2p networks. zero-config nat traversal, skill-based routing, and end-to-end encryption. It connects AI assistants to external tools and data sources through a standardized interface.

How do I install AgentAnycast?

Install via npm with the command: npx -y agentanycast. Then add the server configuration to your AI client's JSON config file (e.g., claude_desktop_config.json or .cursor/mcp.json).

Which AI clients work with AgentAnycast?

AgentAnycast works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is AgentAnycast free to use?

Yes, AgentAnycast is open source and available under the Apache-2.0 license. You can use it freely in both personal and commercial projects.

Browse More Communication MCP Servers

Explore all communication servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "agentanycast": { "command": "npx", "args": ["-y", "agentanycast"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use AgentAnycast?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides