How to plan a multi-stop route with ChatGPT, Claude or Grok

Chat assistants such as ChatGPT, Claude and Grok can help you plan a day with several stops. They are good at turning a messy list into a clear plan. They are not, on their own, routing engines. This guide explains what to expect, how to get a better answer, and when to hand the job to a real routing tool.

What a chat assistant does well

A chat assistant is strong at the parts of route planning that are about language and organization.

Cleaning up a list. It can turn pasted notes, emails or a spreadsheet into one stop per line.
Spotting constraints. It can pull out details like “closes at 5 p.m.” or “pickup at 3:15” that are easy to miss.
Grouping stops. It can suggest grouping stops by neighborhood or town so you do not cross the same area twice.
Drafting a schedule. It can lay out a rough timeline once you give it service times and travel estimates.
Explaining tradeoffs. It can tell you which stop is hardest to fit and why.

What it cannot do alone

A language model works by reasoning in text. Without a connected tool, it has no live road network, no current traffic, and no table of real drive times between your stops. Any distances or drive times it gives you are estimates.

Stop order is also a hard math problem. Choosing the best order for a set of stops is the Traveling Salesperson Problem. Google’s OR-Tools documentation notes that 10 locations already allow 362,880 possible routes, and checking every option quickly becomes “computationally intractable.”

Research shows language models find this hard. The GraphArena benchmark (ICLR 2025) found that even top models “struggle with larger, more complex graph problems and exhibit hallucination issues.” A 2024 study of GPT-3.5 Turbo on the same problem found the share of correct trips declined as the number of stops grew.

In practice, an unaided answer can look sensible but be longer than needed. It can also drop a stop or ignore a time window. That is fine for three errands. It is risky for a work day with fifteen appointments.

How to ask for a better route

You will get a more useful answer if you give the assistant complete, structured input. These tips apply to ChatGPT, Claude and Grok alike.

Give full addresses. Include street, city, region and postal code. Vague names like “the bank” invite guesses.
State where you start and finish. Say whether you return home, end at a depot, or finish somewhere else.
List fixed times separately. Mark appointments and opening hours as hard limits.
Say how long each stop takes. Fifteen minutes at the post office and an hour at a client site change the plan.
Ask it to show its assumptions. Request a list of every guess it made about drive times or locations.
Ask for a stop count check. Have it confirm that every stop appears exactly once.
Ask what would break the plan. A good answer names the tightest window and a fallback.

If your stops come from a spreadsheet, it helps to tidy the data first. Our guide on how to prepare a job list for route planning covers addresses, service times, arrival windows and duplicates, and there is a sample job-list CSV you can copy.

Check the result in a map app

Treat the assistant’s order as a draft, then check it against real road data.

Google Maps lets you add multiple destinations for driving, walking and cycling directions. Its help page says you can add up to 9 stops, including the final destination, and drag stops to change their order. Each route shows an estimated travel time, so you can compare a couple of orders quickly.

Google Maps follows the order you set. It does not promise to find the best order for you. If two orders look close, keep the one that respects your fixed appointments.

Connect a routing tool to your assistant

The bigger upgrade is letting the assistant call a real routing tool. The Model Context Protocol (MCP) is an open standard for connecting AI applications to outside tools and data. The official MCP site lists Claude and ChatGPT among the AI assistants that support it.

Claude. Anthropic’s help center says custom connectors using remote MCP are available on Free, Pro, Max, Team and Enterprise plans, with Free users limited to one custom connector.
ChatGPT. OpenAI’s help center describes developer mode for connecting custom MCP apps, with availability and admin controls that depend on the plan.
Grok. xAI’s documentation says connectors are available to all Grok users and that you can add a custom MCP connector by entering the server URL.

Once a routing tool is connected, the assistant can send your stops to an engine that uses real road data and returns an ordered route with times. Routing companies have started to offer this. For example, Routific’s MCP server, which it describes as in beta, offers a quick tool that optimizes 2 to 8 stops and returns stops in order with ETAs and distances.

When to use a real routing tool

A chat assistant on its own is usually enough for a short personal trip. Move to a routing tool when any of these apply:

You have more than a handful of stops.
Customers have arrival windows you must hit.
You need accurate drive times, not estimates.
More than one driver or vehicle shares the work.
You run the same kind of route every week and want consistent results.

The best setup combines both. The assistant handles the conversation and the messy input. The routing tool does the math. Then the assistant explains the result in plain language.

Where Mappy fits

The tips in this guide ask for the same details a Mappy route request uses. Mappy is an agent skill for route-planning workflows. It brings jobs and driver constraints into a structured request: job addresses and service times, arrival windows and local time zone, and driver starting locations and working hours.

The Mappy site says the skill runs through your agent, in three steps: parse jobs, resolve constraints, and plan the route. The skill package includes agent instructions, input and output contracts, a job-list template and a planning guide.

If you plan multi-stop days with a chat assistant or agent, you can see the Mappy agent package at mappybot.com/download. Start by filling in the sample job-list CSV, using how to prepare a job list for route planning as a checklist.

FAQ

Can ChatGPT, Claude or Grok find the shortest route by themselves?

Not reliably for many stops. Without a connected routing tool they have no live road data, and research such as GraphArena shows language models often give suboptimal or infeasible answers on larger routing-style problems.

How many stops can I add in Google Maps?

Google’s help page says you can add up to 9 stops, including the final destination, for modes other than public transit and flights. You can drag stops to reorder them.

What is the quickest way to improve a chat assistant’s route?

Give full addresses, start and end points, fixed appointment times and service times. Then ask it to list its assumptions and confirm every stop appears once.

Can I connect a routing tool to my chat assistant?

Yes, if the tool offers an MCP server and your assistant supports custom MCP connectors. Claude, ChatGPT and Grok each document a way to add one, with plan-specific rules.

What is Mappy, and what does it cost?

Mappy is an agent skill for route-planning workflows that brings jobs and driver constraints into a structured request. The Mappy site lists a Personal plan at $5 / month, with 10 stops per route, 150 stops per month and single-vehicle route optimization. Business Starter and Business Fleet plans are listed as Custom.

Sources

All guides