Onchain Safety Mcp

io.github.aitools420Published by GitHub account aitools420v0.2.0Updated Oct 24, 2026

Token-safety for AI agents: rug/honeypot verdict + score across PulseChain, Monad, Base & BSC

The publisher's own server: compute.pangle.online — GPU rental market

VerifiedSTDIODesktop onlyFinanceSecurity & Monitoring

Overview

AI-generated overview

Lets an assistant run on-chain token-safety checks — rug/honeypot verdicts, ownership risk, exit liquidity, wallet approvals — on PulseChain, Monad, Base and…

What you can do
A local stdio client over a hosted API that returns deterministic on-chain analysis rather than LLM judgement. Tools include check_token_safety for a verdict and 0-100 score, fresh_rug_radar for the last 20 newly created pools, exit_safety for size-aware price impact, check_ownership for renounce/upgradeable and owner privileges, safe_to_interact as a bundled verdict, plus wallet_approvals and wallet_poison_check for drainer and address-poisoning scans.
When to use it
Use it when an assistant has to judge an unknown token before buying, selling, accepting or interacting with it, especially on PulseChain, which the README says major token-safety APIs do not cover. Also useful for screening fresh pools and auditing a wallet's active approvals.
Requirements
Runs locally via npx (Node.js) as a stdio MCP server; desktop-only, no web executable. Needs network access to the hosted API. The free tier needs no key; an optional ONCHAIN_API_KEY (wsk_ prefix) unlocks the paid deep tier. ONCHAIN_API_BASE can override the API host.
Before you install
Results are informational and not financial advice; treat verdicts as heuristics, not guarantees. The optional ONCHAIN_API_KEY is a secret and should be supplied by environment variable, not committed. The README also describes an x402 pay-per-call route where agents pay USDC on Base per request, which can spend real funds. Queries send chain and wallet or token addresses to a third-party API.

Installation

In SourceWeft

  1. Open Onchain Safety Mcp in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

onchain-safety-mcp

MCP (Model Context Protocol) server that gives any AI agent on-chain token-safety tools for PulseChain, Monad, Base, and BSC — including PulseChain, which the major token-safety APIs (GoPlus, Honeypot.is) don't support at all (verified June 2026). If your agent touches PulseChain, this is the safety check.

It is a thin stdio client over the hosted API at onchain.wick.pics (engine wick-safe/0.3). Every check is deterministic on-chain analysis — no LLM in the verdict path, so results are fast, cheap, and reproducible: contract-risk heuristics, liquidity depth, honeypot transfer-simulation, LP-burn, ownership/renounce + privilege detection, upgradeable-proxy detection, and a wallet approval (drainer) scanner. The free tier needs no key. Informational, not financial advice.

Tools

toolwhat it answers
check_token_safety({ chain, address })Is this token a scam? Verdict (SAFE … LIKELY_RUG), 0–100 score, evidence checks.
fresh_rug_radar()What just launched, and is it safe? Last 20 freshly-created pools, safety-scored at creation.
exit_safety({ chain, token, sizeUsd, maxSlippage? })Could I sell $X of this at acceptable slippage? Size-aware price impact + safety verdict.
check_ownership({ chain, address })Is ownership renounced / upgradeable, and what can an active owner still do (mint, blacklist, pause, tax)?
safe_to_interact({ chain, address })One call → SAFE_TO_INTERACT / CAUTION / DO_NOT_INTERACT, bundling safety + ownership, with reasons.
wallet_approvals({ chain, owner })A wallet's active ERC-20 approvals, flagging unlimited grants — the drainer vector.
wallet_poison_check({ chain, owner })Address-poisoning scan: dust/look-alike transfers in a wallet's history that mimic a real counterparty to trick a paste-and-send of the wrong address.

chain: pulsechain | monad | base | bsc · token/address: 0x… contract address. Agents should call check_token_safety / exit_safety before interacting with, buying, or accepting an unknown token.

Install (any MCP client — Claude Desktop, agent frameworks)

json
{  "mcpServers": {    "onchain-safety": {      "command": "npx",      "args": ["-y", "github:aitools420/onchain-safety-mcp"],      "env": { "ONCHAIN_API_KEY": "wsk_…" }    }  }}

ONCHAIN_API_KEY is optional — omit it for the free rate-limited tier. A wsk_ key unlocks the paid deep tier (get one at onchain.wick.pics).

Pay-per-call for autonomous agents (x402)

The same engine is available with no signup and no key via the x402 pay-per-call protocol — agents pay USDC on Base per request. Machine-readable index of every paid route (token-safety, fresh-rug radar, exit-safety):

https://onchain.wick.pics/.well-known/x402.json

Human docs: onchain.wick.pics/agents

Env

vardefaultpurpose
ONCHAIN_API_BASEhttps://onchain.wick.picsAPI host
ONCHAIN_API_KEY(none — free tier)wsk_ key for the paid deep tier

License

MIT

Source: README.md at commit 3135c8c

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v0.2.0LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

Other Pangle integrations