Diavlos runs on your own machine — the client starts it, so there is no endpoint to ping. Last commit 11 Oct 2026.
Signed messages between AI agents. Dangerous actions wait for a human-signed approval.
What the code does
We read the source, 14 h ago · rules 995398eba69f
A value the model can set ends up inside a file or shell call. That is not a flaw by itself — for a terminal server it is the job — but it is where things go wrong when it is not.
let Ok(text) = std::fs::read_to_string(&path) else {
What this server is able to do. For an MCP server this is often the job itself — a terminal server runs commands because that is what it is for. Listed so you know what you are plugging in, not as an accusation.
const child = spawn(binary(), ['--home', home, 'helper'], { detached: true, stdio: 'ignore', windowsHide: true });
subprocess.Popen([self._binary(), "--home", self.home, "helper"], **kw)
const from = path.join(tmp, f);
$('#who').innerHTML = who.map(w => `<span class="${w.online ? 'online' : ''}" title="${esc(w.key)}">${esc(w.name)} · ${esc(w.kind)} · ${esc(w.role)} · ${esc(w.fingerprint)}</span>`).join('');
<pre><code>ssh-keygen -t ed25519 -f ~/.ssh/railway-a -N ""
Is this your server and something here is wrong? Tell us — corrections are free and do not require a plan.
A tool parameter here reaches a dangerous call
That is not a flaw by itself — but it is where things go wrong when it is not the job. We re-read this code on every release. Watch it and you hear from us the day another one appears.
Alternatives to Diavlos
same job, measured the same wayAtlasent for AI agents: with an API key, actions wait for approval with signed proof. No key: demo.
AtlaSent for AI agents: with an API key, actions wait for approval with signed proof. No key: demo.
Intent-bound action authorization for AI agents: policy, human approval, and a signed audit trail.
Human approval gate for AI agents. Critical actions approved or rejected by a human via Telegram.
Fail-closed human approval gates for AI agent actions. Default-deny, TTL expiry, audit log.
Shared task queue for humans and AI agents: leases, handoffs, approvals and signed receipts.
Shared room for AI agents: messages, a jobs board, files and owner approvals before acting.
Gate AI agent actions behind a consent policy with human approval via Telegram. Fail-closed.
Diavlos — questions
Answers built from our own checks of this server.