The shape here is a workflow: a plan is created for a token, its sections carry a status, and findings accumulate as resources the session can list back. CoinGecko and DefiLlama are addressed by name rather than through a generic fetch, so the two places most token questions end up are each one call away. The tools ship without descriptions, so a client has to choose between them on their names and argument lists alone — which is workable for a set this literally named, and a real cost when two of them differ only in scope.
A token-research workflow server keyed on a token name and ticker, with CoinGecko and DefiLlama addressed as named sources and a plan whose sections carry a status.
- A research plan created from a token's name and ticker, with per-section status maintained through update-status
- CoinGecko reached three ways — a name lookup, a data pull for the token, and a ticker listing bounded by limit
- DefiLlama as the second named source, by search and by token
- Source-scoped work: search-source and research-source both take a source alongside the token, so a question can be aimed at one place instead of the open web
- A second pass driven by terms rather than the plan, through research-with-keywords
- Page content fetched by URL in a chosen format, and list-resources for what the session has accumulated
The token's name and its ticker, since nearly every tool keys on both. No account or key is declared, and the two sources the tools name are CoinGecko and DefiLlama.
One command — npx -y web3-research-mcp@latest