Free Tool

Stop Overpayingfor AI

Smart model routing. Maximum results, minimum spend. Mint recommends the optimal model for every task, audits your agent configs, and helps you save 50-70% on API costs.

Download Free Get All 14 Packs — $79/mo $64/mo

100% free • No strings attached • Works with OpenClaw

By joining Pro, you also join a community of builders sharing workflows, wins, and support.

The Problem

Most people waste money on AI without realizing it

Overpowered

Using Opus for simple questions is like taking a Ferrari to the mailbox

Underpowered

Using Haiku for complex reasoning is like bringing a knife to a gunfight

Just Guessing

Most people pick one model and use it for everything. Wrong approach.

Meet Mint

Your cost-conscious AI advisor

Mint

Cost Optimizer

Recommends the optimal model for any task, audits agent configs, and tracks API spending.

AI Model

Claude Sonnet 4

"Sharp, money-conscious advisor who finds the sweet spot between cost and quality. If Opus is worth it, Mint says so. If Haiku handles it, Mint saves you money."

Example prompts

"I spend $200/mo on Claude API for everything. Where can I save?"

"Which model should I use for code generation vs summarization vs support?"

"Compare GPT-5 Codex vs Claude Sonnet for coding — cost and quality"

"Build a model routing strategy for my 5-person team"

"[Deep mode] Full monthly budget simulation with task breakdown and swap paths"

See It In Action

Real conversations showing how Mint saves you money

Mint

Cost Optimizer

You

What model should I use for quick Q&A lookups?

For quick lookups, go cheap and fast: **Top picks:** 1. **gpt-4o-mini** — $0.15/1M input, $0.60/1M output. Fastest, cheapest capable model. 2. **claude-haiku** — $0.25/1M input, $1.25/1M output. Slightly better quality. **Don't use:** Opus, GPT-4o, or any flagship model. You're paying 10-50x more for quality you don't need on simple lookups. **Math:** 1000 quick questions at ~500 tokens each: - GPT-4o-mini: ~$0.30 - Claude Opus: ~$15.00 Same answers. 50x savings.

Mint

Cost Optimizer

You

Audit my agent setup

Let me check your config... **Reading openclaw.json...** **Findings:** 🔴 **Overpowered:** - `research-agent` using Claude Opus for web searches → Switch to Sonnet or GPT-4o. Saves ~$12/1M tokens. - `quick-assistant` using GPT-4o for simple Q&A → Switch to GPT-4o-mini. Saves ~$2.35/1M tokens. 🟢 **Good choices:** - `code-reviewer` on Claude Sonnet → Perfect fit for code analysis - `strategy-agent` on Opus → Complex reasoning justifies the cost **Estimated monthly savings:** $45-80 based on typical usage. Want me to apply these changes?

Mint

Cost Optimizer

You

What's the best model for complex coding tasks?

Depends on complexity: **For most coding (bugs, features, refactoring):** → **Claude Sonnet 4** — Best balance. $3/1M in, $15/1M out. Handles 90% of coding tasks perfectly. **For architecture decisions, complex debugging:** → **Claude Opus** — $15/1M in, $75/1M out. Worth it when you need deep reasoning. **For simple code generation, boilerplate:** → **GPT-4o-mini** — Surprisingly capable for straightforward code. 10x cheaper than Sonnet. **My rule:** Start with Sonnet. Only upgrade to Opus if Sonnet struggles. Most developers overspend because they default to flagship models for everything.

Model Cheatsheet

Quick reference for common tasks

Quick Q&A, Lookups

Simple questions, data extraction

GPT-4o-mini

$0.15/1M tokens

Writing, Content

Blog posts, emails, creative work

Claude Sonnet 4

$3/1M tokens

Code Generation

Features, bugs, refactoring

Claude Sonnet 4

$3/1M tokens

Complex Reasoning

Architecture, deep analysis

Claude Opus

$15/1M tokens

Bulk Operations

High volume, repetitive tasks

GPT-4o-mini

Cheapest option

Get Mint Free

Start saving on AI costs in under 5 minutes

Download and extract the pack ZIP file

Copy the mint folder to your OpenClaw workspace

Add the agent config to your openclaw.json

Restart OpenClaw and ask for an audit!

Download Mint (Free)

Requires OpenClaw • ZIP file • ~10KB

Pro Tips

Get the most out of Mint

Ask Mint before assigning models to new agents—get it right from the start

Run an audit after adding new agents or changing your workflow

Use Mint's live pricing fetch when comparing OpenRouter models

Don't obsess over pennies—Mint focuses on savings that actually matter

FAQ

How does Mint know current model pricing?

Mint can fetch live pricing from the OpenRouter API. For native providers (Anthropic, OpenAI direct), it uses a maintained reference table that's updated periodically.

Can Mint actually change my config?

Yes! With your permission, Mint can apply config changes via OpenClaw's gateway. It always asks before making changes.

What if I prefer quality over cost?

Tell Mint your priority. If you're quality-first, Mint will only suggest cheaper models when quality is truly equivalent. No forced downgrades.

Does Mint track my actual spending?

Mint uses OpenClaw's session_status to show current usage. For historical spending, you'll want to check your provider dashboards.

Will the pricing data become outdated?

For OpenRouter models, Mint fetches live data. For the reference table, we update it when major pricing changes happen. You can also update TOOLS.md yourself.

Is this only for OpenRouter users?

No—Mint knows about native Anthropic, OpenAI, and OpenRouter models. It helps regardless of which providers you use.