Analysis updated 2026-05-18
Automatically fill in missing category, memo, and receipt information for Ramp card transactions.
Search iMessage and Gmail for a matching receipt and attach it as a rendered PDF.
Run a weekly scheduled sweep so new company card spend stays coded within days.
Flag any transaction where the receipt, category, or memo cannot be confidently determined instead of guessing.
| kothari-nikunj/ramp-autofill | 0-bingwu-0/live-interpreter | 010zx00x1/faresnipe | |
|---|---|---|---|
| Stars | 2 | 2 | 2 |
| Language | Python | Python | Python |
| Setup difficulty | hard | moderate | easy |
| Complexity | 4/5 | 2/5 | 2/5 |
| Audience | ops devops | general | general |
Figures from each repo's GitHub metadata at analysis time.
Requires an authenticated Ramp CLI account plus Gmail and Google Calendar MCP connectors configured in Claude Code.
ramp-autofill is a Claude Code skill that cleans up your Ramp company card transactions automatically. For every transaction it makes sure the category, memo, and receipt are filled in where Ramp requires them, or it explicitly flags the transaction with a reason if it cannot be completed safely. Nothing is silently ignored, every transaction ends up in one of three buckets: already complete, completed during this run, or flagged for you to look at. It finds receipts by searching iMessage for receipt links and searching Gmail, then renders whatever it finds locally into a PDF using a headless browser before attaching it. The safety rules are built into the code itself rather than just suggested in a prompt, so they hold even if something goes wrong elsewhere. Only a small allowlisted set of Ramp commands can ever run, actions like approving payments or moving funds are not in that list at all. The tool also refuses to overwrite a transaction that already has a memo, category, or receipt, and it only attaches a receipt when the amount, date, and merchant plausibly match, otherwise it flags the transaction instead of guessing. The process runs in phases: it first learns your personal categorization style from past transactions, then fetches and sorts transactions by what they need, searches for receipts, has separate passes judge the categorization, memo wording, and receipt matches, and finally writes the results according to whichever automation mode you have chosen. There are four modes, ranging from report only, which changes nothing, up to full auto, which writes both high and medium confidence results without pausing for approval, though low confidence items are never written automatically in any mode. Everything runs on your own machine using the Ramp CLI, Claude Code, and connectors for Gmail and Google Calendar, with no remote services involved.
A Claude Code skill that automatically categorizes Ramp card transactions, finds and attaches matching receipts from iMessage and Gmail, and flags anything it cannot confidently complete.
Mainly Python. The stack also includes Python, Claude Code, Playwright.
The README does not state a license.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly ops devops.
This repo across BitVibe Labs
Verify against the repo before relying on details.