Loading ToolBench
MCP server for one-click ERC20 token issuance and blockchain integration.
Three tools with complete input schemas and descriptions. Naming follows verb_noun pattern (deploy_token, get_token_info, list_deployed_tokens). Descriptions are adequate (50-150 chars) but lack actionable context for LLM selection. Parameters are typed and constrained (e.g., symbol maxLength=11, decimals 0-30), but descriptions are minimal. Output schemas are not documented. Error handling exists but lacks recovery guidance. No tool annotations (readOnlyHint/destructiveHint). Missing pagination for list_deployed_tokens despite potential for large result sets.
Deploy a customizable ERC20-compatible token to the configured EVM network. Supports dry-run estimation.
Fetch metadata for a MintChain-issued token using its contract address.
List tokens deployed via this MintChain MCP instance, including deployment metadata cached locally.
Output schemas not documented. LLMs cannot infer what fields are returned or plan downstream tool calls. deploy_token should document return fields (txHash, contractAddress, deploymentCost, etc.); list_deployed_tokens should document token object structure and total count.
list_deployed_tokens lacks pagination (limit, offset, cursor) and result count. If many tokens are deployed, returning all at once wastes context and risks timeout. Should cap results at 20-50 and offer pagination.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). deploy_token is destructive (writes to blockchain) but not marked. get_token_info and list_deployed_tokens are read-only but not annotated. Agents cannot distinguish safe from risky tools.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Parameter descriptions are minimal. 'recipient' in deploy_token says 'Address receiving the initial supply. Defaults to the deployer wallet.' but does not explain format (0x-prefixed hex, 42 chars). 'address' in get_token_info lacks format guidance. LLMs may pass invalid formats.
Error handling in httpServer.ts and createServer.ts returns generic messages ('Unknown tool', 'Unexpected tool execution error') without recovery guidance. LLMs cannot self-correct or retry intelligently.