yikart/AiToEarn · warning · McpError

ErrorCode.MethodNotFound

ErrorCode.MethodNotFound

Error message

Unknown prompt: ${name}

What it means

The MCP prompts handler received a GetPrompt request whose params.name is not registered in McpRegistryService for this MCP module. The handler looks up registry.findPrompt(moduleId, name) and, when no prompt with that name exists, returns the MCP-protocol error MethodNotFound with 'Unknown prompt: <name>'. This is the standard MCP response for a nonexistent prompt.

Source

Thrown at project/aitoearn-backend/libs/nest-mcp/src/services/handlers/mcp-prompts.handler.ts:65

            : [],
        }))

      return {
        prompts,
      }
    })

    mcpServer.server.setRequestHandler(
      GetPromptRequestSchema,
      async (request) => {
        this.logger.debug('GetPromptRequestSchema is being called')

        try {
          const name = request.params.name
          const promptInfo = this.registry.findPrompt(this.mcpModuleId, name)

          if (!promptInfo) {
            throw new McpError(
              ErrorCode.MethodNotFound,
              `Unknown prompt: ${name}`,
            )
          }

          const contextId = ContextIdFactory.getByRequest(httpRequest)
          this.moduleRef.registerRequestByContextId(httpRequest, contextId)

          const promptInstance = await this.moduleRef.resolve(
            promptInfo.providerClass,
            contextId,
            { strict: false },
          )

          if (!promptInstance) {
            throw new McpError(
              ErrorCode.MethodNotFound,
              `Unknown prompt: ${name}`,

View on GitHub (pinned to d3aa8bea5b)

Solutions

1. Call prompts/list on the connected MCP endpoint and use an exact name from the response
2. Check the registered prompt's metadata.name vs the requested name for casing/spelling differences
3. Verify the client is connected to the MCP module that actually registers this prompt (moduleId/endpoint check)
4. If the prompt should exist, confirm its provider class is registered in the module and not gated behind a disabled feature flag

Example fix

// before (client)
await client.getPrompt({ name: 'summarize_repo' })
// after (name matches registered metadata.name)
await client.getPrompt({ name: 'summarize-repo' })

const { prompts } = await client.listPrompts()
if (!prompts.some(p => p.name === wantedName))
  throw new Error(`Prompt ${wantedName} not offered by this MCP server`)

function promptExists(list: { prompts: { name: string }[] }, name: string): boolean {
  return list.prompts.some(p => p.name === name)
}

try {
  return await client.getPrompt({ name })
} catch (e) {
  if (e instanceof McpError && e.code === ErrorCode.MethodNotFound && e.message.startsWith('Unknown prompt'))
    return await pickPromptFromList(client) // refresh and retry with a valid name
  throw e
}

When it happens

Trigger: An MCP client calls prompts/get with a name that was never registered via @Prompt (or equivalent) on this module, the name is misspelled or differs in case from the registered metadata.name, the prompt belongs to a different MCP module (moduleId mismatch), or the client used a stale prompt list from before the server was redeployed without that prompt.

Common situations: Client caches ListPrompts results from an older server version; a prompt was renamed or removed; multi-module setups where the client connects to the wrong module's endpoint; LLM agents hallucinating prompt names; typos in client config/prompt references.

Related errors

• 任务不存在
• InvalidAiTaskId
• MaterialGroupNotFound
• text prompt is required
• DraftGenerationMemoryNotFound

AI-assisted analysis of yikart/AiToEarn@d3aa8bea5b (2026-08-31). Data as JSON: /api/errors/3e9443ee92326f65. Report an issue: GitHub.