# We just got our travel booking MCP into Claude's official connectors directory. Here's everything the review made us change. — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/2efaeb25b0d6087f

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-10-12 18:41 ingested 2026-10-13 02:37

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1x4a492/we_just_got_our_travel_booking_mcp_into_claudes/)

## retained text



```
Hey guys, I built LetsFG with a friend of mine and our open source community. It's an MCP that lets Claude search and book hotels and flights. Today we got listed live in Claude's official connector directory.

When we started I couldn't find much about what the review actually looks at, so here's everything we had to change, in case you're building a connector too.

1. Tool descriptions must describe, not instruct  
  
Our book\_flight description was like 5,000+ characters of "always do this, never do that", basically steering the model through the description, and that's not allowed. A description just says what the tool does, what it takes and what it returns. So we moved all the "what to do next" stuff into the tool results, and honestly it works better there coz Claude reads it at the exact moment it needs it. We ended up rewriting all 26 tools.

2. Results can't sell  
  
Our search results had lines like "a good deal", which sounds harmless, but Claude repeats the tool results to the user, so you're basically turning Claude into your salesman and nobody wants that, especially Anthropic. Now the results only show the numbers, price, times, seller, reliability etc. and Claude decides on its own if it's a good deal.

3. Spending money needs an email approval, outside of chat  
  
This was a big change. We already had an Approve email, but it only marked the booking as approved, and the next tool call from Claude was what actually put the hold on the card. Anthropic caught that as against the rules, and to be honest it improved our product. Now the booking starts when you press approve in the email, and Claude just watches the progress. Same for hotels and cancellations.   
Small tip if you build smth similar: make the email link open a page with a button instead of doing the action on click, coz email scanners open every link and you'd be booking flights for people just by them receiving the email lol.

4. That means every small charge too  
In the second review they found one more thing, our tool for answering questions during a booking, like the price went up a bit at checkout, there's a paid bag, or a switch to another seller. Claude could just answer "yes" to that from the chat. Their rule: anything that adds or raises a charge needs the same email approval as the booking. Only free stuff like a free seat can stay in the chat. So we had to adjust for this.

5. Guest login was fine as the test account  
Anthropic's review forms ask for a test account with a username and password, but we don't have passwords at all, you connect as a guest or with your email and a code. We told them to connect as a guest and they had no problem with that.

Small things we realized only after doing them:

\- Renaming a tool or changing its description resets every user's "Always allow" for it.  
\- Directory search looks at the name, one-liner, description and tool names.  
\- If you edit your listing after it's approved but before you publish, it goes back to draft and has to be reviewed again. Happened to us, a week long delay :(

How Claude helped: Claude Code did most of the fixes. It rewrote the 26 descriptions, moved the instructions into the results, built the Approve page and the approval emails for price changes, bags and seats. Most useful trick: after the rewrite I ran the same \`claude -p\` prompt against the live server before and after, to check Claude still called the same tools in the same order. A green test suite wouldn't have told me that.

Honestly I have to praise the Anthropic team here. Every note said exactly what to change, re-tests came back within about a day, and everything they flagged made our product safer. We'd have been listed within a week if I hadn't edited the listing. Meanwhile I'm on week 3 waiting for any email from OpenAI :P Thanks Anthropic!

If you're curious this is our connector: [https://claude.ai/directory/letsfg](https://claude.ai/directory/letsfg)

Happy to answer anything about the review.
```



document fetched 2026-10-13 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1x4a492/we_just_got_our_travel_booking_mcp_into_claudes
