Home / Guides / Your Babson AI Key

Your Babson AI Key

On this page
  1. What each tool does
  2. GitHub Student Developer Pack
  3. What you receive
  4. Rules and budget
  5. Set it up in VS Code
  6. Check that it works
  7. Spend the budget well
  8. Where the key is stored
  9. Other agents and your own key
  10. If something breaks

Babson gives you an API key for Claude Sonnet 5, with a $100 monthly budget covered by the College. It is issued for OIM3690, it is meant for your work in this course, and it runs through December 31, 2026.

IT sends the key with a PDF covering Claude Desktop, a separate application for chat and for Cowork. Follow that if you want those. The setup below covers VS Code, which is where the course actually builds things, and it takes about five minutes.

What each tool does#

Tool Use it for Setup
GitHub Copilot Inline completions, the gray ghost text you accept with Tab Getting Started, step 2
Babson AI key Chat and agent mode in VS Code The steps on this page
Microsoft Copilot Chat about anything you would keep out of a consumer account Sign in with your Babson account

Microsoft Copilot runs under Babson's own agreement. It is a chat tool, so it does not edit your files the way an AI agent does.

GitHub Student Developer Pack#

The GitHub Student Developer Pack is free and includes developer tools such as JetBrains IDEs, cloud credits and a free domain name. It also upgrades your Copilot plan to Copilot Student, which makes code completions unlimited and adds a monthly allowance of AI credits for chat and agent mode. GitHub can ask for proof of enrollment, such as a dated class schedule, a transcript, or a student ID, so apply early and leave time for approval.

What you receive#

An email from Babson IT with four values in a table:

Field What it is
Gateway base URL The address your tools send requests to
API key Your credential, starting with sk-
Auth scheme x-api-key
Model The exact model name your tools must ask for, currently claude-sonnet-5

Keep that email. Every step below uses those four values, and this page does not repeat the gateway URL or the key, so read those from your own message.

Rules and budget#

  • Nothing public-facing. Do not put this key in a deployed website, app, or service where other people could trigger requests on it. Your own machine and your own editor are fine. For a project that calls a model from a live site, use your own key from another provider.
  • Do not share it. Every request is attributed to you, so lending the key out means lending your identity.
  • Assume it is logged. Requests through the gateway, including prompts, responses, and tool activity, are recorded and attributable to your key, for cost tracking, support, and institutional records.
  • Use it for coursework. For a personal project, use a key of your own.
  • The budget is $100 a month. It resets at the start of each month, and requests stop when it runs out.

Set it up in VS Code#

  1. Open the Command Palette with Ctrl+Shift+P (Windows) or Cmd+Shift+P (macOS) and run Chat: Manage Language Models.
  2. Click Add Models.
  3. Choose Custom Endpoint.
  4. For the group name, enter Babson AI.
  5. Enter a display name, such as Sonnet 5 (Babson), then paste the API key from your email.
  6. For the API type, choose Messages.
  7. VS Code opens a file called chatLanguageModels.json. Change only its models list. Find the line "models": [, delete everything between that [ and the ] that closes it, and paste the block below in its place. Then replace PASTE_YOUR_GATEWAY_BASE_URL_HERE with the gateway base URL from your email, and save.
{
  "id": "claude-sonnet-5",
  "name": "Sonnet 5 (Babson)",
  "url": "PASTE_YOUR_GATEWAY_BASE_URL_HERE/v1/messages",
  "toolCalling": true,
  "vision": true,
  "maxInputTokens": 200000,
  "maxOutputTokens": 64000
}

Get these right:

  • id must match the Model value in your email exactly, including every hyphen. A near miss returns an error that looks like the key is broken.
  • url is your gateway base URL with /v1/messages on the end. Keep the https:// and add nothing else.
  • Do not touch the apiKey line. VS Code already wrote it, and it looks like ${input:chat.lm.secret...} with an id unique to your machine. It points at the key you pasted in step 5, which VS Code keeps in your system keychain. Copying someone else's line breaks it.
  • Keep the two token values as shown. The deployment accepts up to a million input tokens and 128,000 output tokens. VS Code summarizes the older part of a chat once it nears the input value, so 200,000 also limits how much one request can send.

Save the file. If the model does not show up in the chat model picker, restart VS Code.

Check that it works#

Open the Chat view, click the model picker at the bottom of the input box, and select Sonnet 5 (Babson). Ask it anything. A reply means you are done.

If the model is not in the picker, look for a second dropdown next to it, showing a session target such as Copilot, Claude or Codex. Recent VS Code versions have one, and older ones do not. Set it to Local. A key you added yourself is a custom endpoint, and custom endpoints run under Local, so under any other target the model is missing from the list. That looks like a broken key and is not one.

Know what this key pays for:

  • It powers chat and agent mode, which is where the course spends its AI budget.
  • Inline completions, the gray ghost text you accept with Tab, keep running on Copilot's own free allowance and never touch this key.

Before you start a chat, check the model picker. With Sonnet 5 (Babson) selected, the request draws on your Babson budget. Any other model draws on Copilot's AI credits.

Spend the budget well#

Every message in a chat sends the whole conversation again, including the files and screenshots in it. Each message costs more than the one before, and the replies slow down.

To watch this happen, open Explore the context window. It plays a session in Claude Code, Anthropic's agent, and shows what fills the context window, the text the model reads on every request, as each message arrives. Claude Code is a different agent from the one in VS Code, and both send the conversation and the files they read again with every message.

  • Start a new chat for each task. When a chat gets long, ask the agent to write what you have decided into a Markdown file in the project, then start a new chat and have it read that file.
  • Put several changes in one message. A single message with five changes costs far less than five separate messages.
  • Make small edits yourself. A word, a number or a color is faster to change in the file than to ask for.
  • Paste text you can select, such as an error message, a Console line or a block of code. A screenshot of the same text costs far more tokens, and the agent reads text more accurately. Add a screenshot only when the agent needs to see the screen itself: a layout, a chart, a color.

These habits stretch this key's budget. They are not a reason to use AI less. Token prices keep falling and the models keep improving, so learning to get the most out of the strongest ones is worth what it costs. When a task needs more than this budget allows, use a key of your own and keep going.

Where the key is stored#

chatLanguageModels.json lives in VS Code's user configuration folder, inside your home directory. It sits outside every project folder, so it never lands in an assignment repository and Git never sees it.

Do not paste an API key into a project file, a commit, or a prompt.

Other agents and your own key#

No assignment requires anything in this section. Use it when you already have one of these tools or want to try one.

Claude Desktop#

The PDF that came with your key sets up Claude Desktop, where you get Chat and Cowork. It is a separate application from VS Code and it does not replace the steps above. Set it up if you want it.

Claude Code and Codex#

Claude Code (from Anthropic) and Codex (from OpenAI) are agents with a VS Code extension and a terminal version. Every paid Claude plan includes Claude Code. Codex comes with every ChatGPT plan, including the free one, and which versions you get and how much you can use vary by plan.

Claude Code is switched off on Babson keys by default, because it needs Git configured on your machine. IT can turn it on. Message me if you want it and I will ask on behalf of the class.

A key of your own#

The Babson key is for your own machine and your own editor. Get a key from another provider, such as OpenRouter, in these cases:

  • A live project calls a model. The Babson key cannot go into a website, app or service that other people can use. Anyone who opens a page can read a key inside it, so keep the key in a function on Vercel, as Deploy on Vercel When Your API Needs a Key describes. Set a credit limit on the key, and leave auto top-up off.
  • Your monthly budget runs out. It resets at the start of each month.
  • The Babson key expires on December 31, 2026.

To use the new key in VS Code, run Chat: Manage Language Models, click Add Models, and pick the provider from the list. Then select that model in the chat model picker. Requests on it bill to that key and never touch your Babson budget.

If paying for a key of your own is a barrier, message me.

If something breaks#

What you see Usual cause
The model never appears in the picker The JSON has a syntax error, or VS Code needs a restart
An error naming the model The id does not match the Model value from your email
An authentication error The key was pasted with a stray space, or the URL is missing /v1/messages
It broke right after you pasted into the JSON file Your paste overwrote the apiKey line VS Code wrote. Remove the model under Chat: Manage Language Models, add it again, and paste only the model block
It worked and then stopped The monthly budget ran out, and it resets on the first

For installation and setup problems, email the IT Service Center at support@babson.edu. For anything about the key itself, such as the model, the budget, or a request to raise it, message me first and I will take it to the AI team. Bring everything else to class or office hours.

Last updated: Wednesday, 10/28/2026