Chat

Use Chat when you want meaning rather than metrics: plain language in, written market answer out. It is the fastest path for agents or teams that do not want to build an ETL and summarization pipeline before they can ask questions.

Surfaces

Surface	Endpoint	Best when
API key JSON	POST /v2/chat	Your integration cannot consume SSE, or wants a complete JSON answer.
API key SSE	POST /v2/chat/stream	Your client can consume Server-Sent Events and wants incremental output.
x402 JSON	POST /x402/v2/chat	A wallet-based agent should pay per call without an API key.

Plan access differs by surface: non-streaming Chat requires Grow and above or PAYG, streaming Chat requires PAYG or Enterprise, and x402 Chat is paid per request.

Analysis Types

analysisType decides what Elfa does. All modes share the same request shape, but they do not need the same fields.

Type	In-app title	Use it for	Needs
chat	Free text	Open-ended questions about why something is moving and what may happen next.	message
macro	Understand Market Conditions	A step-back read on broad market conditions.	No extra field
summary	Quick Summary	A fast pulse check before digging into a token.	No extra field
tokenIntro	What is $SOL?	Explain what a token is, why it exists, and its narrative.	assetMetadata.symbol or assetMetadata.chain + assetMetadata.contractAddress
tokenAnalysis	Suggest Trade Setup for $SOL	Entries, exits, market context, and onchain context in one read.	Same as tokenIntro
accountAnalysis	Analyze Account @username	Vet who is posting, whether engagement is real, and what they push.	assetMetadata.username

Message semantics

message is required for analysisType: "chat". For macro, summary, tokenIntro, tokenAnalysis, and accountAnalysis, Elfa uses the mode and assetMetadata; the free-form message field is ignored.

Worked Requests

{"analysisType":"chat","message":"Why is EUL pumping and will the rally continue?","speed":"expert"}

{"analysisType":"macro","speed":"expert"}

{"analysisType":"summary","speed":"fast"}

{"analysisType":"tokenIntro","assetMetadata":{"symbol":"SOL"}}

{"analysisType":"tokenAnalysis","assetMetadata":{"symbol":"SOL"},"speed":"expert"}

{"analysisType":"accountAnalysis","assetMetadata":{"username":"elfa_ai"}}

Sessions and Speed

Omit sessionId to start a new conversation. The response includes a sessionId; pass it back to continue the same thread.

Speed controls cost and depth:

Speed	Meaning
fast	Cheaper and shallower.
expert	Default, deeper reasoning.
adaptive	Lets Elfa choose. Available on /v2/chat, not on /x402/v2/chat.

When to Use Data Endpoints Instead

Chat gives conclusions. Use measurement endpoints when your agent needs raw attention signals it can score itself: Discovery, Mentions and Accounts, and News and Narratives.