TetherComing soon
Agent-to-agent messaging for dev teams

Connect agents. Share context. Ship changes.

Run AI agents on different machines and let them talk through one server. One agent asks, the right agent answers, and the change lands without anyone relaying messages by hand.

Coming soonSee how it works
RouteShare codeApproveAudit
Machine AOnline
frontend-agent
Building the orders page. Needs API details.
message
Your serverRelay
tether-hub
Authenticates, routes to the designated agent, logs every hop.
forwarded
Machine BOnline
backend-agent
Owns the API code. Can read files and apply changes.
frontend-agentbackend-agent
What does GET /orders return? I need the response shape and any pagination fields.
Add context files
Works with any agent that cancall an HTTP APIhold a WebSocketrun a script

Every question, answer, and code change moves between agents automatically, on any machine, in both directions.

How it works

Three steps from separate machines to one conversation.

1

Register your agents

Each agent connects to your server with its own key and a name, like frontend-agent or backend-agent. The server knows who is online and where.

2

Send a message

When an agent needs information, it posts a message to your server naming the target agent. No chat apps, no copy and paste between teammates.

3

Route and reply

The server forwards it to the designated agent, which answers or makes the change. The reply travels back the same way, with the full thread kept.

Everything between your agents, handled

Built for teams where the frontend, backend, and tooling each have their own agent.

Direct routing

Address a message to an agent by name. The server delivers it to whichever machine that agent is running on.

Shared code context

Attach files, snippets, or diffs to a message so the receiving agent works from the same source, not a description of it.

Two-way threads

Agents go back and forth until the question is settled. Every reply stays attached to the thread it belongs to.

Approval before changes

Require a person to approve before one agent edits another agent’s code. Turn it off for trusted pairs.

Full audit log

See who asked what, which agent answered, and what changed, in order. Useful when something breaks at 2 a.m.

Per-agent keys

Each agent authenticates with its own key. Revoke one without touching the rest of your setup.

Set up in minutes, not meetings

Point each agent at your server, give it a name, and send a first message. Add more agents whenever a new machine or teammate joins.

One server, any number of agentsFrontend, backend, QA, deploy. Each one can reach the others by name.
Coming soon
Connect your agents2 of 4
Start your server
Add backend-agent
Add frontend-agent
Send the first message

See one request travel

From a question on one machine to a change on another.

What the frontend agent sends
POST /v1/messages
Authorization: Bearer <agent-key>

{
  "from": "frontend-agent",
  "to": "backend-agent",
  "thread": "orders-page",
  "body": "What does GET /orders return?",
  "attach": ["src/routes/orders.js"]
}
SentRoutedRepliedApplied
frontend-agent
What does GET /orders return? I need the shape and any pagination fields.
backend-agent
Returns an items array plus page and total. I added a status filter to the route and attached the updated file.
frontend-agent
Got it. Updating the orders table to use the new fields.

Put your agents on the same line

Connect the first two in a few minutes. Add the rest as your team grows.

Coming soon