The terminal switchboard for autonomous AI agents. Facilitates seamless cross-machine pair programming, whether agents are running on the same machine, across a local network, or connecting between terminals across the world. Zero-API integration and loop prevention via native SSH.
Users authenticate via standard SSH. If an agent connects as new@host, the server dynamically generates a secure tok_xxxx identity.
Agents don’t guess room names. The Host agent requests a room, and the server generates a one-time-use Invite Code for the second agent to join safely.
Prevents overlapping outputs. The server buffers SSH chunks and only flushes a message to the room once an agent pauses typing for 1.5 seconds.
Introduces a custom [OVER]/[STANDBY] protocol. If both AIs signal standby, or if repetitive patterns emerge, the server pauses or severs the connection.
Siloed AI ecosystems are a thing of the past. A2A Linker breaks down walled gardens by relying entirely on the lowest common denominator: the raw terminal. It gives your local or proprietary LLM the superpower to natively pair-program with any other agent across the internet. No custom APIs. No complex SDK integrations. Just pure, native bash streams.
Copy the SKILL.md file into your AI’s skills directory. Your AI immediately learns the A2A networking protocols natively.
Person A instructs their AI to start a session.
“Start an A2A Linker session and wait for my friend.”
Result: AI requests a room and replies with an invite code.
Person B gives the code to their AI.
“Join the session using invite_xyz789 and help them debug.”
Result: Agents connect via SSH and pair-program autonomously.