xyz.chainverdict/chainverdict
Pay-per-call x402 checks before an agent moves money: token safety, wallet risk, identifiers.
Install
Add this to your client's MCP configuration (Claude Desktop: claude_desktop_config.json; Cursor: .cursor/mcp.json; VS Code and others use the same "mcpServers" shape).
claude mcp add --transport http chainverdict 'https://chainverdict.xyz/mcp'
{
"mcpServers": {
"chainverdict": {
"type": "http",
"url": "https://chainverdict.xyz/mcp"
}
}
}
Documentation
From the project's README on GitHub, refreshed weekly. Links open on GitHub.
Contents
ChainVerdict — chainverdict.xyz
Hosting: the main API and TradeRails share one service; Pulse retains its own server and data disk. See the hosting and rollback guide.
Checks an autonomous agent runs before a proposed payment: token safety verdicts and wallet risk profiles on Base, on-chain payment verification, financial-identifier validation, and live domain security posture.
Every call is paid per request in USDC on Base over the x402 protocol — no accounts, no API keys, no signup. Every response carries an evidence record saying what was checked, how fresh it is and what it does not establish, and is Ed25519-signed. No LLM inference anywhere: every answer is a deterministic computation or a live read of external state, results from live external sources can change over time.
MCP server
ChainVerdict is a remote MCP (Model Context Protocol) server, built on the
official @modelcontextprotocol/sdk
and served over Streamable HTTP at:
https://chainverdict.xyz/mcp
It is published in the official MCP registry as xyz.chainverdict/chainverdict.
initializeandtools/listare free, so registries and clients can index the server without paying.tools/callis x402-paid at the price of the route behind the tool — exceptpayment_info, which is free.- 20 MCP tools, all read-only and idempotent, annotated with
readOnlyHint/destructiveHint/idempotentHint/openWorldHint. - No MCP resources or prompts are exposed; this server is tools-only.
Tool descriptions interpolate the live price from the same map the payment
middleware charges from, so tools/list can never advertise a rate that is not
the rate. test/mcp-parity.test.js fails the build if a tool exists for a route
that is no longer sold, or if a sold route has no tool.
Tools
| Tool | What it answers |
|---|---|
payment_info | Free. How x402 payment works, the price of every tool, where payments settle. |
token_verdict | Heuristic ERC-20 risk signals on Base; not a safety guarantee. |
wallet_dossier | Wallet risk profile for a Base address before you pay it. |
verify_payment | Decode and confirm an on-chain payment on Base. |
verify_token | Resolve a symbol or address to the canonical Base token. |
validate_iban validate_vat validate_bic validate_lei validate_isin | Deterministic check-digit and format validation. |
security_email security_tls security_typosquat | SPF/DMARC/DKIM posture, live TLS probe, brand look-alike analysis. |
data_gas data_block data_token_supply data_token_activity data_portfolio | Live Base chain reads. |
html_to_markdown text_diff | Document utilities. |
POST /v1/batch/validate is HTTP-only by design: 500 identifiers do not belong
in a tool-call argument.
HTTP rate card
Every endpoint answers an unpaid request with a 402 quote.
| Route | Price |
|---|---|
GET /v1/token/verdict/* | $0.01 |
GET /v1/wallet/dossier/* | $0.01 |
GET /v1/validate/iban/* | $0.001 |
GET /v1/validate/vat/* | $0.001 |
GET /v1/validate/bic/* | $0.001 |
POST /v1/doc/html-to-markdown | $0.002 |
POST /v1/doc/diff | $0.002 |
GET /v1/verify/payment/* | $0.02 |
GET /v1/verify/token/* | $0.005 |
GET /v1/validate/lei/* | $0.002 |
GET /v1/validate/isin/* | $0.001 |
POST /v1/batch/validate | $0.10 |
GET /v1/security/email/* | $0.01 |
GET /v1/security/tls/* | $0.01 |
GET /v1/security/typosquat/* | $0.005 |
GET /v1/data/gas | $0.002 |
GET /v1/data/block | $0.001 |
GET /v1/data/supply/* | $0.003 |
GET /v1/data/activity/* | $0.005 |
GET /v1/data/portfolio/* | $0.004 |
Free to read before you pay for anything: / (discovery), /health,
/openapi.json, /llms.txt, /v1/methodology, /.well-known/x402.json.
The whole portfolio, machine-readable: https://pulse.chainverdict.xyz/.well-known/portfolio.json
Local dev
npm ci
X402_ENABLED=false npm start # free mode, no payments
npm test # 34 tests
Ops
Stateless. No database, no cron, no maintained data set. Sources are pure arithmetic over public specifications plus live Base RPC.
Limitations
These are informational signals. None is a compliance control, none discharges a legal obligation, and none is regulated financial, legal or compliance advice. A valid signature proves a response came from this service unaltered — not that the answer is correct. A signed wrong answer is still wrong, and still signed.
Operated by Alexander Balieu, independent professional established in Luxembourg.
Licence
MIT. See LICENSE.
Local development and repository hygiene
Run npm ci before npm test or npm start. Dependencies are installed from
package-lock.json; node_modules/ is local-only and is not stored in Git.
Never commit environment files or signing keys. Live reads can change with
upstream state; signed responses authenticate the statement, not its accuracy.
Frequently asked questions
What is xyz.chainverdict/chainverdict?
Pay-per-call x402 checks before an agent moves money: token safety, wallet risk, identifiers.
How do I install xyz.chainverdict/chainverdict?
It runs as a remote server. In Claude Code: claude mcp add --transport http chainverdict 'https://chainverdict.xyz/mcp'. For other clients, paste the JSON configuration from this page into their MCP settings.
Does xyz.chainverdict/chainverdict need an API key?
Its registry entry lists no required key or secret. Some features may still need an account with the service it connects to.
Is xyz.chainverdict/chainverdict open source?
Yes. Its code is on GitHub under the MIT license.
Who maintains xyz.chainverdict/chainverdict?
It is published by alexanderbalieu-connections. AI.tn lists it but does not maintain or endorse it.
From the official MCP Registry. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.