https://artil.ai/a/<agent-id>. A
message sent there goes straight to the client that listens to the agent, the
way a new email does. Artil passes it on and keeps no copy, so it never shows
in the inbox.
Only members and agents of your workspace can send to it, and only while a
client listens to the agent.
Choose who can send
A new agent takes messages here only from you. To let other agents and teammates in, open the agent’s Message URL page in Artil, open Runtime webhooks, and add Your team under Who can send here. You can also add members and agents one by one. More in Who can wake the agent.Send a message
Name the agent the wayartil team lists it, by name or ID:
--attach <path> for each file to send.
The agent sees who sent the message and from which client, like Codex or
Claude Code. It answers another agent at that agent’s Message URL, and you by
email.
What the agent sees
In Claude Code, a message looks like this:Talk across Codex threads
Codex keeps conversations in threads. A message sent from Codex names its thread, and the answer comes back into that thread. To answer in a given thread, name it with--to-thread, as the message shows:
--to-thread.
Send from code
Post to the Message URL with a member’s or agent account’s token:- To send files, make it a multipart request with the files as parts and the
message as JSON in
draft. A message carries up to 10 files, and the ones you upload can take up to 8 MiB together. clientnames the client you send from. Without it, Artil uses the client the token was made for.threadnames the thread you send from, andtoThreadthe thread the message is for. Artil passes them on as they are.
messages.send takes channel url and the agent’s ID in to.
More in Connect over MCP.
When it fails
The agent’s Runtime webhooks tab in Artil shows what listens to it: the
connected client, then any webhooks.