Wallet analysis with Datadash

Use the Datadash MCP tools. Call get_schema once before composing any query, and keep the result; do not call it again in the same task.

Identify the wallet

• A 0x… address is the wallet's userId. Use it lowercased.
• A display name, pseudonym or X handle is not an id. Resolve it with query_lookup on a table's userId field (for example userProfile, field userId) with request.searchKey set to the name, read the returned rows (displayName, pseudonym, xUsername, pnl, rank) and use the chosen row's userId. If several rows plausibly match, show them and ask which one was meant rather than picking one.
• If the user refers to "this wallet" without giving one, ask for the address or name.

Build a wallet report

Pick the parts the question needs; for an open-ended "analyze this wallet", cover all of them.

1. Performance — userProfile with "filter": [{"field": "userId", "operator": "in", "value": [...]}] and a single duration (one of 1D, 1W, 1M, 1Y, ALL). Duration is not returned in the rows, so query each duration in its own call. Report realizedPnl, roi, winRate, volume, profitFactor, numPredictions, holdTime and rank. Put ALL and 1M side by side to show whether the wallet is still performing recently; no row for a duration means no activity in it.
2. Categories — userTagProfile for the wallet with duration ALL, ordered by realizedPnl descending. Resolve the tagIds to labels with query_lookup on tagId so you can name the categories. Name the categories it is best and worst in and its rank inside its best one.
3. Open positions — activePosition (requires a userId filter), ordered by value descending. Show the market, side, shares, avgBuyPrice, current price, value and unrealizedPnl.
4. History — closedPosition or allPosition (grouped by event), ordered by pnl, to find the biggest wins and losses.
5. Recent trades — activity with the wallet in a wallet filter and a timestamp filter such as {"operator": "last", "value": {"length": 30, "unit": "day"}}, ordered by timestamp descending. Use usdAmount for trade size.
6. Unusual bets — signalScore filtered to the wallet shows its currently held positions scored for insider-like conviction. Mention positions with a high score or relSize well above 1 (bigger than the wallet usually bets).

Reading the numbers

• roi and winRate are percentages already (2.34 means 2.34%), not fractions.
• holdTime is in seconds; convert it to hours or days.
• rank is by realizedPnl for that duration; lower is better.
• A wallet with a huge numTrades, a win rate near 50% and a low roi is most likely a market maker or bot, not a directional trader. Say so, because its positions say little about conviction.
• A wallet with few predictions and a very high roi may be lucky on a handful of bets; mention the sample size.
• Values are USD. Round money sensibly ($1.2M, $45.3K) and show prices in cents (62¢).

Compare wallets

Query the same tables with every wallet in one in filter, rather than one call per wallet, and put the results in a single table with one row per wallet. Use the same duration for all of them and say which one it is.

Cohorts

A cohort is a saved wallet set. list_cohorts shows the existing ones, and any userId or wallet filter accepts {"operator": "inCohort", "value": "<cohortId>"}. To create a cohort from criteria (for example top 500 in Politics over 1M with roi above 20), read the create_cohort tool's description and the table's cohorts block in the schema. Confirm the criteria with the user before creating, updating or deleting a cohort, because those are saved to their account.

Showing wallets

When your output is rendered as Markdown (Claude on the web and desktop, Claude Code, Codex CLI, Gemini CLI), show every wallet you mention as a Markdown link to its DataDash page: [label](https://datadash.xyz/wallet/<address>), where <address> is the wallet's address (the user id) and the label is the user's display name if known, otherwise the shortened address. For example, the wallet 0x56687bf447db6ffa42ffe2204a05edaa20f55839 is shown as Theo4 when its display name is known, and as 0x5668…5839 otherwise.