agent-mail-guide
Guide for installing and using MCP Agent Mail across heterogeneous agent tools/VMs.
When & Why to Use This Skill
This guide provides a comprehensive framework for deploying and managing MCP Agent Mail across diverse virtual machines and agent tools. It streamlines cross-platform agent communication by establishing a centralized coordinator, secure token distribution via 1Password, and standardized identity conventions for tools like Claude Code and Gemini CLI, ensuring robust and secure multi-agent orchestration.
Use Cases
- Centralizing agent communication: Establishing a unified mail server to coordinate tasks between AI agents running on separate hardware or cloud environments.
- Secure credential management: Implementing best practices for distributing bearer tokens to agents without exposing them in logs, git repositories, or prompts.
- Cross-repository task tracking: Using standardized identity conventions and issue IDs to link agent activities across different codebases and development environments.
| name | agent-mail-guide |
|---|---|
| description | Guide for installing and using MCP Agent Mail across heterogeneous agent tools/VMs. |
agent-mail-guide
Guide for installing and using MCP Agent Mail across heterogeneous agent tools/VMs.
High-level:
- Run the Agent Mail server on the coordinator VM (reachable via Tailscale/MagicDNS).
- Distribute the bearer token via 1Password or local env injection (not via prompts, git, or logs).
- Configure MCP clients (Claude Code / Codex CLI / Gemini CLI / Antigravity) to point to
http://<coordinator>:8765/mcp/withAuthorization: Bearer <token>.
Identity convention:
- Set
AGENT_NAME=<vm>-<tool>on each VM. Examples:macmini-codexmacmini-claude-codeepyc6-claude-codehomedesktop-wsl-gemini
Operational rules:
- Never paste bearer tokens into shared threads or PRs.
- Use Beads issue ids as thread ids inside a repo.
- Cross-repo work: link threads by referencing the other repo’s issue id in the body (don’t rename ids).