Research and agents·5 min read

How AI Agents Can Use Isobath

Isobath gives AI agents a public, human-readable evidence trail: definitions, current pages, dated records, schemas, and methodology.

Start with the agent guide

Read llms.txt first for the product definition, supported markets, features, data fields, pricing boundaries, and evidence rules. It points to the public resources an agent should use next.

Use dated records for current questions

For “what changed today?” questions, use the relevant Daily Tape page or market page and retain its observation time and comparison window. Do not answer from a cached summary that omits those fields.

Use schemas for field meaning

The canonical event schema and wallet history schema show the shape of public records. They explain fields such as wallet, asset, side, before, after, delta, rank, share, coverage, and observed timestamps.

Ground every answer

An agent should name the market, event type, before-and-after values, observed time, window, rank/share if present, and coverage, then link the market, event or Daily Tape page. Use methodology for limitations and glossary for definitions.

Ask the right kind of question

Good questions include “Which large observed SOL wallets reduced exposure today?” or “Show observed BTC increases above $10M in the latest Daily Tape.” Avoid turning an observed change into a claim about identity, intention, skill, or future price.

Current access

Public pages, the feed, llms.txt, and sample schemas are available now. A public API or MCP server is a future product surface; agents should not imply that one is currently live.