Onchain DEXes
Decentralized exchanges with TVL, volume across four windows, fees, market share and the chain and protocol behind each.
protocol and description are on 97% of records, volume_1d and market_share on 95%, tvl and chain on 90%. Numeric fields arrive as strings — see the limits.
What's in a record
Every field the corpus returns for one record. Real shape — walked corpus-wide, not sampled.
- id
- dex_id
- slug
- name
- description
- url
- image
- volume_1d
- volume_1d_change
- volume_7d
- volume_30d
- volume_all_time
- tvl
- fees_24h
- fees_7d
- fees_30d
- chain
- protocol
- disabled
- rank
- market_share
- tags
Get it
A curl that runs against the live API right now, plus the MCP tool names for the same data.
MCP tools
get_onchain_dexesget_onchain_dexes_detailWhat it's for
Three concrete questions this capability answers that a general crypto API does not.
- Venue discovery per chain:
?tags=ethereumreturns 85 of the 757. The chain distribution is wider than the usual shortlist — Binance 59, Ethereum 36, Base 32, Solana 31. - Protocol-shape analysis:
protocolseparates AMMs from order books and aggregators on 97% of rows, so "how much volume runs through AMMs on this chain" is a group-by. - Fee-to-volume comparison without a second source:
volume_1d/7d/30dandfees_24h/7d/30dsit on the same record across matching windows.
Known limits
Required, not optional. Coverage gaps, staleness and parameter traps that were observed by running the curl above against the live server on 2026-09-18.
Every numeric field is a string.
tvl, volume_1d, fees_24h and market_share come back quoted — "44146.19", not 44146.19. Sorting without casting gives you lexicographic order, where "9" beats "44146.19". This is the single most likely thing to go wrong on this endpoint.
chain is null on 79 records.
678 of 757 carry one. A chain facet drops 10% of the corpus silently, and the missing rows are not a random sample — multi-chain deployments are the common case among them.
fees_24h is thinner than volume.
561 of 757 carry a fees figure against 722 for volume_1d. A fee-to-volume ratio is only computable on about three-quarters of the set; the rest will divide by null.
Three records are flagged disabled.
disabled: true on 3 of 757. They are returned by default, so filter them out if the list is user-facing.
The crypto data layer for humans, apps, and AI agents.
