Beta
Discord Download

ZenMux vs AutoRouter

For Claude Code and Codex on a Mac, AutoRouter is the shorter route: it picks a model for each turn, uses the Claude and ChatGPT plans you already pay for first, and needs no API keys. ZenMux is a hosted LLM gateway with one API for 100+ models, model routing, and insurance that compensates for poor output or high latency. This page covers how ZenMux works, then where each one fits.

AutoRouter A Mac app that picks a model for each turn. $0/mo on your plans Download ZenMux A hosted gateway with one API for 100+ models. $20/mo Starter, or pay as you go
Price $0 on your plans, Pro $50 a month, Enterprise by seat Pay as you go from a $5 top-up, or a Builder Plan at $20, $100 or $200 a month
Fee on tokens Provider’s price plus 10% on paid models Per-token rates listed on each model’s page
Runs as A Mac app, macOS 14 or later A hosted API on Cloudflare’s edge network
Agents Codex, Claude Code, Cursor, Claude Desktop, T3 and Terminal Claude Code, Codex, Claude Desktop, OpenCode, Cline and more
Setup Install, sign in, click Enable Create a ZenMux key, then set it in each agent’s environment or config
Your plans Claude, ChatGPT, Grok and Gemini plans, used first ZenMux credits, or a ZenMux Builder Plan
Model choice Picked for each turn, after your rules The model you name, or zenmux/auto per request
Teams MDM rollout, allowed models, budgets, SSO, SCIM Organizations with roles, a shared balance and limits per member
Requests go Through AutoRouter’s servers to each provider Through ZenMux’s gateway to providers or their cloud partners
  • Uses the plans you already pay for
  • A model picked for each turn inside your agent
  • No API keys to create or paste
  • Rollout to every Mac in a team through MDM

Sources checked November 17, 2026: ZenMux home page, ZenMux introduction, Quickstart, Claude Code guide, Codex guide, Model Routing, Insurance, Subscription Plans, Pay As You Go, Organizations and AutoRouter pricing. AutoRouter wrote this page. See every comparison.

How ZenMux works

What the gateway is, how it connects to coding agents, how its routing picks a model, and what its insurance covers.

One API for 100+ models

ZenMux calls itself an enterprise LLM platform with one API for 100+ models from OpenAI, Anthropic, Google, DeepSeek and others. It speaks the OpenAI protocols at zenmux.ai/api/v1 and the Anthropic protocol at zenmux.ai/api/anthropic, and names models as provider/model (Quickstart). Requests go to the nearest node on Cloudflare’s edge network. It also has a single fallback model per request, provider routing, request logs, cost and usage pages, and a Studio for chat, image and video.

There are two ways to pay. Pay As You Go is a prepaid balance, topped up from $5, with no rate limit. The Builder Plan is $20, $100 or $200 a month for “personal development, learning, and Vibe Coding,” with rolling five-hour quotas in Flows and 10 to 15 requests a minute (Subscription Plans).

How setup works

Sign up, pick Pay As You Go or a Builder Plan, and create a key. For Claude Code, the guide sets ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN in your shell profile, clears ANTHROPIC_API_KEY, and maps other models through the ANTHROPIC_DEFAULT_*_MODEL variables. For Codex, it exports ZENMUX_API_KEY and adds a ZenMux provider to config.toml, which the CLI and the app both read.

Model routing and insurance

Set the model to zenmux/auto and ZenMux picks a model for each request from your list, or from its full pool, weighing the prompt, task type, price and availability against a balanced, performance or price preference (Model Routing).

Its insurance covers, in its own words, “poor performance, hallucinations, and excessive latency.” Checks run daily on call data and payouts are credited the next day. Compensated cases are anonymized and given back to you as examples to improve your own product (introduction).

What it does well

  • One key reaches 100+ models over the OpenAI and Anthropic protocols, so you skip separate provider accounts.
  • It works from any system and suits your own apps and servers as well as coding agents.
  • Its insurance compensates you for poor output or high latency, paid as credits the next day.
  • Routing picks per request, with a preference you set and a pool of models you choose.
  • Organizations get Admin and Member roles, a shared balance, credit limits per member and usage by member.
  • Pay As You Go keys can carry their own credit, rate and model limits.

Where it gets harder

  • It runs on ZenMux credits or a Builder Plan. Its Claude Code and Codex guides use a ZenMux key and don’t use Claude or ChatGPT plans.
  • The guides set a model by name, so in your agent you pick the model unless you set up zenmux/auto yourself.
  • Each developer edits a shell profile or config.toml on each machine, and the guides list caveats, such as Claude Code not loading settings.json variables after a logout.
  • Builder Plans are for non-production use, with 10 to 15 requests a minute and Flows whose dollar value can change.
  • The insurance page lists Unsatisfactory Content and High Latency but not the thresholds that trigger a payout.

How AutoRouter differs

ZenMux is an API you point your tools at and pay per token. AutoRouter lives inside your agent on a Mac, uses your plans first and picks the model for every turn.

The plans you already pay for

You connect Claude, ChatGPT, Grok and Gemini plans, and they go to the work that needs them, so they last longer. You can use them in any harness, such as a Claude plan in Codex. Free connects two subscriptions and Pro connects any number. When one account reaches its limit, AutoRouter moves to the one with the most room left before it spends any credit. A subscription isn’t required: Pro gives every model, with $20 of included usage each month.

A model for each turn

AutoRouter shows up as one more model in the picker you already use. Your plain-English rules go first, such as “Use Claude Opus 5.5 for UI work.” Otherwise it works out how hard each turn is and only considers models that clear the quality bar for it. A task that continues keeps its model, or moves up. Every turn shows its model, what it cost and why.

Rolled out to a team’s Macs

Enterprise rolls AutoRouter out to every Mac through Jamf or another MDM, with nothing for anyone to set up. Admins get allowed models, team budgets and zero data retention in one place, spend by team, project or ticket, and SSO, SCIM and an audit log.

What it costs and what it can’t do

Routing across your plans is free. Pro is $50 a month with $20 of included usage, and paid models bill at the provider’s price plus 10%. AutoRouter runs only on macOS 14 or later, doesn’t serve your own apps’ API traffic, and has no insurance payouts. It sometimes misjudges how hard a task is. See privacy for what is kept.

Which one to choose

Choose ZenMux if you want one API key for your own product and agents on any system or server, want compensation when output or latency falls short, or need image and video models through the same gateway.

Choose AutoRouter if you work in Claude Code or Codex on a Mac, already pay for Claude or ChatGPT plans, and want each turn sent to the right model with no keys or config files. For other options, read Model routers for Claude Code and Codex, compared.

Setup, side by side

The steps each one’s own docs give, as of November 17, 2026.

ZenMux

  1. Sign up Log in with email, GitHub or Google, then pick Pay As You Go or a Builder Plan.
  2. Create a key Make an API key on the management page for your plan.
  3. Connect your agent Set the base URL and key in your shell profile for Claude Code, or add a ZenMux provider to Codex’s config.toml.
  4. Choose models Map models by name, or call zenmux/auto with a model pool.

AutoRouter

  1. Download AutoRouter for macOS, signed and notarized by Apple.
  2. Sign in Sign in or create a free account, then sign in to your plans.
  3. Enable your apps Allow Accessibility and click Enable for each app.
  4. Pick AutoRouter In a new session, choose AutoRouter from Codex’s model menu, or with /model in Claude Code.

Questions

What people ask about ZenMux.

What is ZenMux?

A hosted LLM gateway that describes itself as an enterprise LLM platform. One ZenMux key reaches 100+ models over the OpenAI and Anthropic protocols, with model routing, fallbacks, usage analytics and insurance payouts.

How much does ZenMux cost?

Pay As You Go is a prepaid balance, topped up from $5, spent at each model’s per-token rate. Builder Plans are $20, $100 or $200 a month for non-production use, and a free plan covers Studio Chat only.

What is ZenMux’s model insurance?

ZenMux says it compensates you automatically for poor output, hallucinations and excessive latency. Checks run daily on call data, payouts are credited the next day, and compensated cases come back to you anonymized.

Does ZenMux work with Claude Code and Codex?

Yes. Its Claude Code guide sets the base URL and a ZenMux key as environment variables in your shell profile, and its Codex guide adds a ZenMux provider to config.toml for the CLI and the app.

Can ZenMux use my Claude or ChatGPT plan?

Its Claude Code and Codex guides use a ZenMux key, paid from ZenMux credits or a Builder Plan. AutoRouter uses your Claude, ChatGPT, Grok and Gemini plans first.

Does ZenMux pick a model automatically?

With zenmux/auto, it picks a model for each request from a pool, by your cost or quality preference. Its coding-agent guides set models by name. AutoRouter picks a model for each turn inside your agent, after your rules.

How is AutoRouter different from ZenMux?

ZenMux is a hosted API you pay per token and configure each tool to call. AutoRouter is a Mac app that needs no API keys, uses your plans first, includes models, picks one for each turn, and rolls out to a team’s Macs through MDM.

How do I move over from ZenMux?

Install AutoRouter, sign in to your plans and click Enable for each app. Enabling adds AutoRouter to the app’s own settings after saving a copy of the file. Rewrite any routing preference you rely on as plain-English rules.

What does AutoRouter cost?

Routing across your ChatGPT and Claude plans is free. Pro is $50 a month with $20 of usage, and paid models bill at the provider’s price plus 10%. See pricing, or estimate a month with and without AutoRouter in the savings calculator.