Screening a single wallet is a question; screening a treasury's counterparties is a job. Returning an id and a signature straight away, then polling, is the right shape for that — and it keeps a slow analysis from blocking a conversation. Treat the tool counts on their own site with care: three pages give three different numbers, and the fourteen here comes from the page that actually lists them.
ChainAware's prediction endpoint, served over SSE. Fourteen tools by their index page, covering wallet risk, token analysis and — unusually — trust scores for registered AI agents.
- predictive_fraud — a fraud probability and AML signals for a wallet before you transact with it, with a batch variant
- predictive_behaviour — a wallet's on-chain history profiled into predicted next actions and a segmentation, also batched
- predictive_rug_pull — risk for a contract or liquidity pool, cross-checked against the deployer
- token_rank_list and token_rank_single — tokens ranked by holder-community strength
- credit_score — a 1-9 trust score from behaviour, fraud probability and social graph
- run_token_audit and get_token_audit_result — a multi-module contract audit, polled every 5 seconds
- agents_trust_score_list and agents_trust_score_single — ERC-8004 registered agents with 0-1000 trust scores
- check_job_status and get_job_results for the batch jobs
An API key in an X-API-Key header, from your ChainAware profile. Their own pages give conflicting tool counts elsewhere — the setup guide says ten, the navigation says six — so the fourteen here is from the index that enumerates them.
Paste a URL with a key — add the endpoint to your client