Tools

Machine-native tools.

Amounts come back scaled and symbolised, counterparties labelled where known, and every field whose text is controlled by the transaction’s author is named in provenance so an agent never treats it as an instruction. Strict JSON in and out, priced per call, built to be called by agents rather than browsed by humans.

base-transaction-decoder

live

Deterministic blockchain transaction decoding for autonomous AI agents.

Structured JSON output
No LLM in the response path
What happened: action, assets, counterparties
Risk analysis with evidence
Agent-ready responses (MCP + x402)

v0.2.0 Base mainnet $0.02 per call 50 free calls a day per IP (IPv6: /64)

Validated on 100 random live Base transactions before launch, 95 decoded to a specific action type, the rest to an honest partial, zero crashes.

Open playground Documentation GitHub

explain_transactionexcerpt

{
  "summary": "0x401d…f2c5 swapped 0.03 ETH
    for 12,899,422 WNL via Uniswap V4
    PoolManager.",
  "action_type": "swap",
  "status": "success",
  "counterparties": [
    { "label": "Uniswap V4 PoolManager" }
  ],
  "risk_flags": [
    { "flag": "unverified_contract" }
  ],
  "gas_paid_usd": 0.020562
}

Excerpt of an actual response for a live Base transaction, trimmed for width. Full schema in the docs.

What it decodes

Swaps on Uniswap V2, V3, and V4 and Aerodrome; ERC-20, ERC-721, and ERC-1155 token transfers; NFT mints, transfers, and Seaport sales; OP-stack bridge deposits and withdrawals; Aave V3 and Compound V3 lending; staking, wrapping, and claims; ERC-4337 account-abstraction bundles; EAS attestations; Basename registrations; and contract deployments, 30 action types from roughly 40 builtin event formats, with app-specific events named via verified ABIs on Sourcify. When a transaction can’t be fully decoded, the output says so instead of guessing. Full enum in the response schema, common questions in the FAQ.

More tools are in development. New releases are announced on the changelog and on GitHub.