top of page

Decision profiles
inside Claude (MCP)

Paste someone's LinkedIn About section (or any other information you have about them) into Claude and ask how they make decisions. Then hand it your draft and ask it to rewrite. Installs by double-clicking a file, and there's no spreadsheet to keep.

WIZE SNAPS API  ·  RECIPE 02  · MCP FOR CLAUDE

What it looks like
 

You're in an ordinary Claude conversation. You paste what the person has written about themselves and ask the question. Claude calls the Wize Snaps API, gets the decision profile back, and tells you what it means for what you're about to send.

 

Example:  The prospect is a Chief People Officer who wrote, in her own words, that she'd rather take an extra fortnight and bring the room with her than be right on her own in week one.

 

Wize Snaps read her as a Collaborator with a strong Analyzer secondary, at High confidence.

 

The draft she was about to be sent said: "Hi Priya, quick one. We can lift decision quality by 30% and we can have you live inside a fortnight. Worth 15 minutes on Thursday to lock it in?"

 

What came back:

 

"Live inside a fortnight" is the worst line in it. She told you she'd rather take an extra fortnight and bring the room with her. You've offered speed to someone who sees speed as the cost of skipping consultation.

 

"30%" is exactly the transformation-on-a-slide claim she's already said she distrusts. With no source and no comparable organisation, the Analyzer side of her will discount it and then discount you.

 

Every criticism is tied to something she actually wrote. That's the difference between this and generic advice about tone.



Why not just ask Claude
 

Ask Claude on its own and it guesses, and the guess can change each time you ask. With the extension, the read comes from the API, with a confidence level and the reasoning quoted from the person’s own words. Same input, same answer.

​


SETTING IT UP
 

  • Download the extension. [Download Wize Snaps for Claude →] 
     

  • Double-click the file. Claude Desktop opens and asks for your Wize Snaps API key. [Get a key →] New accounts start with 100 free credits.
     

  • Turn it on. This is the step people miss. After installing, go to Settings, then Extensions, click Configure on Wize Snaps, and check the toggle says Enabled. It doesn't switch itself on.

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

 

  • Restart Claude Desktop. Quit it properly, not just close the window. The extension reads your key when it starts, so it won't see a key you added mid-session.
     

  • Ask it something. Paste a person's About section and ask how they make decisions.
     

If you ever change your API key, restart again. Same reason.

Screenshot 2026-09-24 at 3.11.22 pm.png


If you use something other than Claude Desktop
 

​The same server works in Claude Code, Cursor, and anything else that speaks MCP. The repo has the config to paste in. Perplexity supports local MCP servers on their Mac app for paid subscribers, so it runs there too, though the setup is more manual.

What does it costs
 

New accounts get 100 free credits. A decision profile costs 1 credit and a rewrite costs 2, so a person fully read and written to costs 3.

​
Two things it won't do

​

Read a LinkedIn URL. The API reads the text you give it and doesn't fetch pages. If you paste a link, you'll get a profile inferred from the name in the URL, which looks exactly like a real answer and isn't one. The extension spots this and tells you.

 

Write your email for you. The rewrite comes back as a strong draft with square-bracket placeholders where it needs a date, a name or a reference. It also doesn't check whether the claims in your message are true. In the example above, the API's own rewrite kept the unsupported 30% in a message aimed at someone who'd said she distrusts unsupported claims. Read what you send.

​

Build on it
 

The server is a couple of hundred lines with no dependencies, and it's on GitHub along with everything else we've built on this API. Read it, fork it, or use it as a worked example for wiring the two API calls into something else.

 

[See the code →] 

 

The whole API is two POST requests with one header. Anything that can make an HTTP call can use it.

​

​

​

​

Is there an MCP server for Wize Snaps?
Yes. Wize Snaps publishes a free MCP server that gives Claude two tools: reading how a specific person makes decisions, and rewriting a message to suit them. It installs in Claude Desktop by double-clicking a .mcpb file, and works in Claude Code, Cursor and other MCP clients via a config entry.

What does the Wize Snaps MCP server do?
It calls the Wize Snaps decision profile API. Given a person's name and text they have written about themselves, it returns one of seven decision profiles, a secondary profile, a confidence level and the reasoning behind the read. A second tool takes a draft message and returns what works, what is likely to land badly for that person, and a rewrite.

Does it work with ChatGPT or Perplexity?
Perplexity supports local MCP servers on its Mac app for paid subscribers, so it runs there with manual setup. ChatGPT requires a remote MCP server, which Wize Snaps does not currently offer.

What it cost?
The server is free and open source under an MIT licence. It calls the Wize Snaps API, which gives new accounts 100 free credits. A decision profile costs 1 credit and a message rewrite costs 2.

Do I need to write code to use it?
No. Download the .mcpb file, double-click it, paste your API key, and switch the extension on in Claude Desktop's settings.

Where is the source code?
github.com/developer-wizer/wize-snaps-recipes

bottom of page