Harness runs on your own machine — the client starts it, so there is no endpoint to ping. 31 installs a week from npm. Last commit 24 Oct 2026.
Start Claude Code and Codex agents the person can watch, wait for them and read their answers.
What the code does
We read the source, 23 h ago · rules 995398eba69f
Things with no honest explanation: a promise that contradicts the code, code that runs at install time while hiding what it does, data leaving the machine.
`rm -rf ${next}`,
&& curl -fsSL https://deb.nodesource.com/setup_22.x | bash - \
"$T/etc/sudoers.d/30-harness-updates" "$T/usr/lib/harness-os/init-keyring"
'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLDAxNDQ0Hyc5PTgyPC4zNDL/2wBDAQkJCQwLDBgNDRgyIRwhMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjL/wAARCAABAAEDASIAAhEBAxEB/8QAHwAAAQUBAQEBAQEAAAAAAAAAAAECAwQFBgc…
shadow = next((line.split(':')[1] for line in Path('/etc/shadow').read_text().splitlines()
What this server is able to do. For an MCP server this is often the job itself — a terminal server runs commands because that is what it is for. Listed so you know what you are plugging in, not as an accusation.
'/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLDAxNDQ0Hyc5PTgyPC4zNDL/2wBDAQkJCQwLDBgNDRgyIRwhMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjL/wAARCAABAAEDASIAAhEBAxEB/8QAHwAAAQUBAQEBAQEAAAAAAAAAAAECAwQFBgc…
if (!::AttachConsole(ATTACH_PARENT_PROCESS) && ::IsDebuggerPresent()) {
const res = await pub.eval(
execFileSync(process.execPath, [prismaCli, 'db', 'push', '--schema', schema, '--skip-generate'], {
child = subprocess.Popen([node, worker, os.ttyname(slave)], stdin=subprocess.PIPE,
…alled${dsh.linked ? ' (linked)' : ''}${dsh.commit ? ` @ ${dsh.commit.slice(0, 8)}` : ''}${update.updateAvailable ? ` · update available (${update.availableCommit!.slice(0, 8)})` : ''} · ${dsh.dir}`)
await Process.run('/bin/chmod', ['700', root.path]);
let xattr = Process()
const file = path.join(runtimeDir, 'pty', version, SERIAL_EXE)
RUN curl -fsSL https://harness.autonomous.ai/cli/install.sh | bash
"just long enough": {"codex":"/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx…
$('eggArt').innerHTML = html;
hn = run(folder / 'harness-tui', '--version', env=dict(os.environ, HN_AS_TMUX='0'))
root = next(row.split(':') for row in Path('/etc/shadow').read_text().splitlines() if row.startswith('root:'))
docker run --rm --privileged --platform linux/arm64 \
for path in folder.iterdir() if folder.is_dir() else []:
Is this your server and something here is wrong? Tell us — corrections are free and do not require a plan.
We found things in this code
Code changes quietly between releases, and nobody reads the diff of a dependency. We do, on every release — watch Harness and you get told the day something new turns up.
Connect this server
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add harness -- npx -y @autonomous-ai/harness-mcp{
"mcpServers": {
"harness": {
"command": "npx",
"args": [
"-y",
"@autonomous-ai/harness-mcp"
]
}
}
}[mcp_servers.harness]
command = "npx"
args = ["-y", "@autonomous-ai/harness-mcp"]{
"mcpServers": {
"harness": {
"command": "npx",
"args": [
"-y",
"@autonomous-ai/harness-mcp"
]
}
}
}{
"servers": {
"harness": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@autonomous-ai/harness-mcp"
]
}
}
}Alternatives to Harness
same job, measured the same wayA portable persona for MCP clients — the same agent in Claude Code, Codex, Cursor and VS Code
Persistent terminals and one launcher for Claude, Codex, Grok, and Cursor harnesses.
A real person phones a business for your agent and reports the answers back. Ten calls a month
Start, prompt and read AI coding agent sessions (Claude Code, Codex, Gemini CLI and more) in Codeman
Persistent memory for AI agents: one brain shared across Claude, Cursor, Codex and more.
Shared, project-aware memory for Claude Code, Codex, and other agents with human review.
Phone numbers for AI agents: order a private number, read the SMS verification code, hand it back.
A package manager for Agent Skills — discover, vet, and pin them across Claude Code and Codex.
Harness — questions
Answers built from our own checks of this server.