mukundakatta avatar

Agentvet

mukundakatta/agentvet-mcp

Wraps the agentvet validation library as three MCP tools you can call from Claude Desktop, Cursor, or any MCP client. validate_tool_args checks an arguments object against a shape spec and returns an LLM-ready retry message when something's missing or wrong. lint_tool_definition scans a tool's JSON schema for common mistakes that hurt LLM accuracy, like missing descriptions or non-snake_case names. generate_retry_message builds the canonical feedback string from a validation error. Useful when you want to audit a batch of tool definitions or prototype validation logic in chat before wiring it into your agent runtime. For production validation loops, the underlying agentvet library runs synchronously in your Node process and throws ToolArgError directly.

GitHub stars
1
tools
3
transport
STDIO
last updated
Apr 27, 2026

Tools

3 tools

Public tool metadata for what this MCP can expose to an agent.

validate_tool_argsCheck args against a small shape spec; returns { valid, error?, retry_hint? } where retry_hint is a ready-to-send LLM feedback message.

Check args against a small shape spec; returns { valid, error?, retry_hint? } where retry_hint is a ready-to-send LLM feedback message.

No parameter schema in public metadata yet.

lint_tool_definitionSanity-check a tool definition for common mistakes that hurt LLM tool-use accuracy.

Sanity-check a tool definition for common mistakes that hurt LLM tool-use accuracy.

No parameter schema in public metadata yet.

generate_retry_messageGiven a validation error, build the canonical LLM-facing retry message using agentvet's ToolArgError.toLLMFeedback() formatting.

Given a validation error, build the canonical LLM-facing retry message using agentvet's ToolArgError.toLLMFeedback() formatting.

No parameter schema in public metadata yet.

Sponsors

Readme

agentvet-mcp

MCP server for @mukundakatta/agentvet. Lets Claude Desktop, Cursor, Cline, Windsurf, Zed, or any other MCP client validate LLM-generated tool-call args before execution and produce LLM-friendly retry messages when something's wrong.

npx -y @mukundakatta/agentvet-mcp

Three tools:

  • validate_tool_args — check args against a small shape spec; returns { valid, error?, retry_hint? } where retry_hint is a ready-to-send LLM feedback message.
  • lint_tool_definition — sanity-check a tool definition for common mistakes that hurt LLM tool-use accuracy.
  • generate_retry_message — given a validation error, build the canonical LLM-facing retry message using agentvet's ToolArgError.toLLMFeedback() formatting.

Add to your client

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "agentvet": {
      "command": "npx",
      "args": ["-y", "@mukundakatta/agentvet-mcp"]
    }
  }
}

View the full README on GitHub

Details

Repository
github.com/mukundakatta/agentvet-mcp
Website
github.com/MukundaKatta/agentvet-mcp
Category
AI & LLM Tools
Package
@mukundakatta/agentvet-mcp
Transport
STDIO
Registry
MCP registry, active

More from mukundakatta

Related AI & LLM Tools MCP servers

View all
1new1direction avatarKorg Ledgernew1direction/korgexVerifiable cognition: prove a session ledger is tamper-evident; audit & import any agent's logs.1stars
2agentra-commerce avatarXapagentra-commerce/xap-sdkAgent settlement protocol — discover, negotiate, settle, and verify with XAP.0stars
3pquattro avatarMemoraEUpquattro/memoraeu-memoraeu-mcpZero-knowledge persistent memory for Claude — encrypted, semantic search, EU hosted.0stars
4u2n4 avatarVEO 3.1u2n4/gemini-media-mcp4K video generation via Google VEO 3.1: text-to-video, image-to-video, video extension0stars
5ivanantigravity-lgtm avatarModel Council Mcp Serverivanantigravity-lgtm/model-council-mcp-serverThree-model MCP server returning Grok, Gemini, and DeepSeek outputs for Claude.1stars
6jayvee6 avatarApple Mail MCPjayvee6/apple-mail-mcpLLM access to Apple Mail and iCloud via AppleScript. Read, search, compose, reply, flag and delete.1stars