Our engineers have been building AI tools like CRAZY the past month...

@tryramp
Ramp@tryramp
21 views Mar 27, 2025 ~1 min read
Advertisement
1
Our engineers have been building AI tools like CRAZY the past month — already saving the whole team hundreds of manual hours.

Our mission is to save businesses time & money. That drive for efficiency starts with our culture.

Here are some highlights 🧵
2
Introducing Ramp’s MCP server, built by Junho Choi on our API team.

We turned our API into an SQL interface that developers can now query with @AnthropicAI's Claude — going from struggling over hundreds of data points to accurately analyzing tens of thousands of spend events.

Run complex analysis on data with a single sentence and build integrations by using natural language, or by delegating the integration entirely to AI agents and LLMs using SQL.


0:47
@josiah_arcanist
Josiah Parappally@josiah_arcanist
🔥TALK TO YOUR FINANCES🔥

In a few days, we went from idea -> open source on an MCP that lets you talk to your financial data like you'd chat with an expert. Ask questions, generate reports, find savings using natural language.

Check out the blog post 👇
engineering.ramp.com/ramp-mcp
3
Speaking of MCPs, Zack built an In-Context Learning with Postgres MCP.

This tool instructs the LLM in-context on how to handle "transaction aborted" errors so it will rollback as necessary to recover from them.


@zack_field
Zack@zack_field
In-Context Learning with Postgres MCP

We've found that in-context learning is a powerful way to teach agents how to better leverage tools.

Here we instruct the LLM in-context on how to handle "transaction aborted" errors and it will rollback as necessary to recover from them.
Media image
4
Rakesh built this tool using Claude Code to gather context from @incident_io and pinpoint problematic code paths.

+ it works with Zack's MCP to create network effects where LLM agents can use multiple MCPs to replicate our manual incident triaging flows.


@rakesh_nori
Rakesh Nori@rakesh_nori
How I built an incident triaging MCP in a few hours with Claude code to save @tryramp on-call engineers time.

It automates context gathering from incidents and pinpoints the codepaths featuring the issues quickly, making incident response fast & efficient.
Media image
5
@incident_io Alex built this Video Process Mining tool that converts Ramp process videos into descriptive maps in order to automate tasks.

Upload a video, ask a question, then the tool processes it with an LLM to provide a process map for engineers to dig from there.


@shevchenkoaalex
shevchenko@shevchenkoaalex
Excited to share an internal tool I've been working on at @tryramp: Video Process Mining.

The idea was to make documenting workflows easier. You just upload a screen recording and it uses AI to generate a visual process map + explanation of the video.

We work with complex financial processes. Now anyone on our team can better understand user walkthroughs and other workflows.
6
@incident_io Roland built this AI Copy Editor that's saving our UX writing team countless hours, all while keeping everything on-brand.

His @figma plugin, AI code reviewer, and codebase auditor automatically enforce our style guide across design and engineering.


@rolandtshen
Roland Shen@rolandtshen
I spent 3 hours in Cursor developing internal design tools for @tryramp product/eng/design teams, saving hours per week.

At Ramp, we ship quickly — but keeping our UX copy consistent and on-brand is a challenge. Our small team of UX writers can’t review every design, and it’s difficult for engineers, PMs, and designers to follow best practices.

To solve this, I worked on a set of AI tools, including an AI-powered Figma plugin, to enforce our style guide and content standards 🧵
Media image
Actions
What You Can Do
  • Export as PDF or Markdown
  • Batch Export to Notion
  • Bookmark & Highlight
  • Screenshot Tweet
Create Free Account

Includes 7-day Premium trial

Advertisement