predictive_fraudLABSCO SUMMARY
Risk scores with a stated accuracy figure and a way to pay without a subscription.
Two things separate this from a generic block explorer wrapper. First, the predictions are published with backtesting numbers — 98% for fraud detection, 90.1% for rug-pull detection — and each result carries forensic detail rather than a bare score. Second, x402 support means an agent can pay per call and never hold a subscription key, which suits one-off checks. Batch work is two-step by design: schedule the job, poll check_job_status, and only fetch results once it reports completed.
What it is
A hosted Web3 risk service reached over SSE. It answers questions about blockchain addresses — likelihood of fraud, what an address is likely to do next, whether a pool looks like a rug pull — across ETH, BNB, POLYGON, TON, BASE, TRON and HAQQ.
What you get
predictive_fraudscores one wallet for fraud probability and runs AML checks;predictive_fraud_batchqueues the same job for a list of addressespredictive_behaviourprojects what an address is likely to do next and profiles its history;predictive_behaviour_batchdoes it in bulkpredictive_rug_pullflags liquidity pools and contracts likely to rugcredit_scorefolds fraud probability, inflow/outflow analytics and social-graph signals into a riskRating from 1 (highest risk) to 9 (highest trust)token_rank_listandtoken_rank_singlerank tokens by the strength of their holders, with search, filter, sort and paginationrun_token_auditreturns a completed audit if one exists or queues a new one;get_token_audit_resultpolls for the reportagents_trust_score_listandagents_trust_score_singlescore ERC-8004 registered AI agents 0-1000 from on-chain behaviour rather than votescheck_job_statusreturns completed/failed/pending counts andget_job_resultsreturns the finished addresses
Requirements
A remote SSE endpoint at https://prediction.mcp.chainaware.ai/sse with a ChainAware API key in the X-API-Key header, from a plan at chainaware.ai/pricing. An x402-compatible client can pay per call instead: the server answers HTTP 402 with payment details and the client settles automatically.
Setup effort
One command plus a key — claude mcp add --transport sse chainaware-behavioural-prediction-mcp-server https://prediction.mcp.chainaware.ai/sse --header "X-API-Key: your-key-here", then supply credentials
USE CASES
01
Vetting an address before sending funds
Score the wallet for fraud probability and read the forensic breakdown behind the number.
02
Warning users away from a risky pool
Check a contract for rug-pull likelihood before anyone deposits into it.
03
Underwriting a DeFi loan
Turn a wallet into a single trust number your lending logic can threshold on.
04
Screening a CSV of addresses
Queue the whole list, poll for progress, and collect the finished addresses when the job reports completed.
05
Deciding whether to call another agent
Look up an ERC-8004 agent's trust score, derived from what it has actually done on-chain.
14 tools
Listed in project docs. Your client may show a different list.
predictive_fraud_batchpredictive_behaviourpredictive_behaviour_batchpredictive_rug_pullcredit_scoretoken_rank_listtoken_rank_singlerun_token_auditget_token_audit_resultagents_trust_score_listagents_trust_score_singlecheck_job_statusget_job_resultsBEFORE YOU INSTALL
What to know first
Tools you get
14 tools named in the project's documentation. Access is by request with a private backend, so without a plan or an x402-capable client none of the tools can be called. The two tools ending in _batch return a job_id immediately; check_job_status and get_job_results fetch results.
Requires
- A client that can reach a remote MCP server over SSE
- A ChainAware API key from chainaware.ai/pricing, in the X-API-Key header (or apiKey in the URL) — or an x402-compatible client instead
INSTALLATION
Connect your client
Server setup
Setup
ChatGPT, Claude.ai
Nothing to build or host — this is a remote SSE server at https://prediction.mcp.chainaware.ai/sse
Get access: subscribe at https://chainaware.ai/pricing for an API key, or plan to use an x402-compatible client and skip the key entirely
Claude Code (CLI): claude mcp add --transport sse chainaware-behavioural-prediction-mcp-server https://prediction.mcp.chainaware.ai/sse --header "X-API-Key: <YOUR_API_KEY>"
Cursor / Windsurf: add the mcpServers entry shown in the Cursor tab to mcp.json
Claude Web / Claude Desktop and ChatGPT: Settings → Integrations (ChatGPT: Apps/Connectors → Add Connector) and enter the URL with the key in the query string:
ENDPOINT URL
https://prediction.mcp.chainaware.ai/sse?apiKey=<YOUR_API_KEY>Optional extra, only if you want the bundled Claude Code subagents: git clone https://github.com/ChainAware/behavioral-prediction-mcp.git then cp -r behavioral-prediction-mcp/.claude/agents/ <YOUR_PROJECT>/.claude/agents/ and export CHAINAWARE_API_KEY="<YOUR_CHAINAWARE_API_KEY>" — this must come AFTER the server is registered, since each agent's frontmatter references the registered server name
Hosted endpoint
ENDPOINT URL
https://prediction.mcp.chainaware.ai/sseAdd this hosted endpoint to your MCP client using the credentials listed under Requirements.
MCP Clients
Claude Code
CLAUDE CODE · TERMINAL
claude mcp add --transport sse chainaware-behavioural-prediction-mcp-server https://prediction.mcp.chainaware.ai/sse \
--header "X-API-Key: <YOUR_API_KEY>"Cursor
CURSOR · JSON
{
"mcpServers": {
"chainaware-behavioural-prediction-mcp-server": {
"url": "https://prediction.mcp.chainaware.ai/sse",
"headers": {
"X-API-Key": "<YOUR_API_KEY>"
}
}
}
}
