Guides

Build a token research report in seconds

Pull price, liquidity, holders, X, web and GitHub data in parallel, let an AI rate the token from 1 to 10 and keep every report in your own table.

Before you buy a token you want five answers: is there real liquidity, who holds the supply, is anyone talking about it, what does the web say, and is anybody still building. Each answer lives behind a different API with its own key, rate limit and format. Collecting them by hand takes ten minutes. On a Vellum sheet it takes seconds, and the AI gets all of it at once.

What you build

One sheet that takes a token mint, pulls five data sources in parallel, lets an AI rate the token from 1 to 10 and saves the result as a row in your own table. It never trades, so you can run it on anything.

Parts

  • Start to run the sheet on demand
  • Token Info for price, market cap and liquidity
  • Holder Check for the top holders and how concentrated the supply is
  • Twitter Check for the project's X account and recent posts
  • Web Search for what the web says about it
  • GitHub for recent development activity
  • AI Assistant to weigh everything and write the verdict
  • Database to keep every report

Steps

  1. Place Start on the left of an empty sheet at /tools.
  2. Add the five data parts in a column to its right and wire Start into each of them.
    • Token Info: paste the mint address.
    • Holder Check: the same mint, mode Top 20.
    • Twitter Check: the project's handle without the @.
    • Web Search: the token name plus "solana".
    • GitHub: the repository URL from the project's site.
  3. Add AI Assistant and wire all five data parts into it. Its prompt:
    Rate this token from 1 to 10. Weigh liquidity, holder concentration, social activity and development activity. Answer with a score and one short line per factor.
  4. Add Database, create a table called token_reports and wire AI Assistant into it. Every run adds one row.
  5. Press Run. Watch the five sources fill in at the same time. The data arrives in parallel, so the whole report is ready in seconds.

Reading the result

The AI answers with a score and one line each for liquidity, holders, social and development. The row in token_reports keeps the time, token, score and top-10 share, so after a week you have your own research log to compare against.

If one wallet holds a large share, check whether it is a liquidity pool. Holder Check also reports the shares without pools, and pools at the top of the list are not a risk on their own.

Make it yours

  • Swap Start for Schedule to re-check a watchlist every hour.
  • Add a second AI Assistant that only answers "buy", "watch" or "skip", then feed that into a Blocker if you ever want to act on it.
  • Share the token_reports table publicly so others can read your research.

Use the same data from your own agent

Every part on this sheet is also an HTTP endpoint. Create an API key on the Agent API page, then call for example POST /agent/holdercheck with {"address": "<mint>"} and the header Authorization: Bearer <your key>. The same parts are available as MCP tools at /mcp, so Claude, ChatGPT or Cursor can pull this data themselves.

Parts used in this guide

Written by Vellum team @useVellum
Follow on X

Build this sheet

Lay out the same parts on the drafting table and run it in preview before it trades.

Open the drafting table