Complete Setup Guide — March 2026

Your AI agent is
burning🔥🔥🔥🔥 your money.
It doesn't have to.

You pay for a Claude subscription. You also pay per-token for the API. This guide eliminates the API bill entirely.

Get the Guide Here
$ moto-api
 
moto-api v1.0.0
 
Backends detected:
✓ claude (7 models)
 
Server: http://localhost:3456
Key: moto_sk_7f3a...9e2b
 
Endpoints:
POST /v1/chat/completions OpenAI format
POST /v1/messages Anthropic format
GET /v1/models
 
✓ Your agent can't tell the difference.
$0
Monthly API cost
45m
Setup time
7
Models routed
The Problem

You're paying for Claude twice

Your Claude Pro subscription gives you access through the CLI. But your AI agent needs an HTTP API — so you're also paying Anthropic per-token. Every message, every cron job, every Slack response burns API credits on top of your subscription.

Without this guide

  • $20/mo subscription + API costs
  • Per-token billing adds up fast
  • Agent retries multiply costs
  • Cron jobs burn tokens 24/7
  • $50–500+/mo in API fees

With this guide

  • $20/mo subscription only
  • Zero API token costs
  • Same models, same quality
  • Agent doesn't know the difference
  • Works with any API-compatible app
Security & Safety

No OAuth. No account bans.

Other tools use OAuth, which has gotten accounts flagged and banned. Zero Token uses simple API key auth with the official Claude CLI. No OAuth, no session hijacking, no risk. Your subscription stays secure.

How It Works

A local proxy your agent can't distinguish from the real API

Your agent sends a standard API request. The proxy translates it into a CLI command and streams the response back. From your agent's perspective, nothing changed.

Your AI Agent
│
standard API request
▼
Moto API — local proxy on your VPS
│
claude -p "your prompt"
▼
Claude CLI → your subscription
Anthropic API — $0.015/1K tokens  →  $0.00
What's Inside

9 parts. Every command. Every fix.

This isn't a blog post with screenshots. It's a production-tested playbook — every step was debugged on a live VPS running real agent workloads.

01

VPS & Docker Setup

Fresh Ubuntu server to Docker-ready in 5 minutes. Node.js, Docker Compose, prerequisites.

02

Claude CLI — Install & Auth

Install the CLI, authenticate with your subscription, fix the 0-byte binary issue, and set up permissions for Docker.

03

The Proxy — Install & Configure

Install the proxy tool, test it manually, and verify it detects your CLI backend.

04

Docker Compose — The Full Stack

Production docker-compose.yml with every volume mount explained. Server patch to reroute API calls through the proxy.

05

Slack Integration

Create a Slack app, configure socket mode, connect it to your agent. Fix duplicate messages.

06

Recommended Settings

Default to the cheapest model, disable background tasks, optimize for subscription limits.

07

Troubleshooting — Every Error We Hit

Auth mismatches, gateway failures, spawn errors, rate limits, memory spikes, sandbox restrictions — with exact fixes for each.

08

Security

What's exposed, what's not, firewall rules, and what to never do.

09

Usage Limits & Cost Optimization

Understand your subscription caps, reduce usage, and monitor consumption.

Built From Pain

Not a tutorial. A postmortem.

Every bug fix in this guide exists because we hit it in production. Every troubleshooting step exists because we wasted hours before finding the answer.

This guide was written after 12+ hours debugging a real production setup — fixing auth mismatches, container networking, CLI quirks, and agent integration issues one by one. You get the clean, working version.

FAQ

Questions

Does this work with Claude Pro ($20/mo)?
Yes. Claude Pro, Claude Max ($100 and $200 tiers) all work. Higher tiers get higher usage limits.
Is usage actually free?
You pay $0 in API token fees. Usage counts against your subscription's CLI limits. For most agent workloads, this is far more cost-effective than per-token billing.
Does this only work with OpenClaw?
No. The proxy speaks standard OpenAI and Anthropic API formats. Any app that expects an API endpoint can use it — Cursor, Continue, custom apps, anything.
What VPS do I need?
Minimum 8GB RAM, Ubuntu 22.04+. A $20–40/mo VPS from any provider works fine.
Do I get updates?
Yes. The guide and proxy tool are updated when bugs are found or when the CLI changes. You get access to all future versions.
What if something breaks?
The troubleshooting section covers every error encountered during real production use — auth failures, timeouts, memory issues, sandbox restrictions, and more. Each has an exact fix.

Stop burning tokens.

One guide. One tool. Zero API costs.

$29
one-time · lifetime updates · includes proxy tool
Get the Guide Here