Filemedium importancesource

mcp.ts

entrypoints/mcp.ts

197
Lines
6270
Bytes
1
Exports
18
Imports
10
Keywords

What this is

This page documents one file from the repository and includes its full source so you can read it without leaving the docs site.

Beginner explanation

This file is one piece of the larger system. Its name, directory, imports, and exports show where it fits. Start by reading the exports and related files first.

How it is used

Start from the exports list and related files. Those are the easiest clues for where this file fits into the system.

Expert explanation

Architecturally, this file intersects with mcp. It contains 197 lines, 18 detected imports, and 1 detected exports.

Important relationships

Detected exports

  • startMCPServer

Keywords

tooltoolsutilsserveroutputschemanameconvertedschematoolusecontexttoolpermissioncontextzodtojsonschema

Detected imports

  • @modelcontextprotocol/sdk/server/index.js
  • @modelcontextprotocol/sdk/server/stdio.js
  • @modelcontextprotocol/sdk/types.js
  • src/state/AppStateStore.js
  • ../commands/review.js
  • ../commands.js
  • ../Tool.js
  • ../tools.js
  • ../utils/abortController.js
  • ../utils/fileStateCache.js
  • ../utils/log.js
  • ../utils/messages.js
  • ../utils/model/model.js
  • ../utils/permissions/permissions.js
  • ../utils/Shell.js
  • ../utils/slowOperations.js
  • ../utils/toolErrors.js
  • ../utils/zodToJsonSchema.js

Source notes

This page embeds the full file contents. Small or leaf files are still indexed honestly instead of being over-explained.

Open parent directory

Full source

import { Server } from '@modelcontextprotocol/sdk/server/index.js'
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'
import {
  CallToolRequestSchema,
  type CallToolResult,
  ListToolsRequestSchema,
  type ListToolsResult,
  type Tool,
} from '@modelcontextprotocol/sdk/types.js'
import { getDefaultAppState } from 'src/state/AppStateStore.js'
import review from '../commands/review.js'
import type { Command } from '../commands.js'
import {
  findToolByName,
  getEmptyToolPermissionContext,
  type ToolUseContext,
} from '../Tool.js'
import { getTools } from '../tools.js'
import { createAbortController } from '../utils/abortController.js'
import { createFileStateCacheWithSizeLimit } from '../utils/fileStateCache.js'
import { logError } from '../utils/log.js'
import { createAssistantMessage } from '../utils/messages.js'
import { getMainLoopModel } from '../utils/model/model.js'
import { hasPermissionsToUseTool } from '../utils/permissions/permissions.js'
import { setCwd } from '../utils/Shell.js'
import { jsonStringify } from '../utils/slowOperations.js'
import { getErrorParts } from '../utils/toolErrors.js'
import { zodToJsonSchema } from '../utils/zodToJsonSchema.js'

type ToolInput = Tool['inputSchema']
type ToolOutput = Tool['outputSchema']

const MCP_COMMANDS: Command[] = [review]

export async function startMCPServer(
  cwd: string,
  debug: boolean,
  verbose: boolean,
): Promise<void> {
  // Use size-limited LRU cache for readFileState to prevent unbounded memory growth
  // 100 files and 25MB limit should be sufficient for MCP server operations
  const READ_FILE_STATE_CACHE_SIZE = 100
  const readFileStateCache = createFileStateCacheWithSizeLimit(
    READ_FILE_STATE_CACHE_SIZE,
  )
  setCwd(cwd)
  const server = new Server(
    {
      name: 'claude/tengu',
      version: MACRO.VERSION,
    },
    {
      capabilities: {
        tools: {},
      },
    },
  )

  server.setRequestHandler(
    ListToolsRequestSchema,
    async (): Promise<ListToolsResult> => {
      // TODO: Also re-expose any MCP tools
      const toolPermissionContext = getEmptyToolPermissionContext()
      const tools = getTools(toolPermissionContext)
      return {
        tools: await Promise.all(
          tools.map(async tool => {
            let outputSchema: ToolOutput | undefined
            if (tool.outputSchema) {
              const convertedSchema = zodToJsonSchema(tool.outputSchema)
              // MCP SDK requires outputSchema to have type: "object" at root level
              // Skip schemas with anyOf/oneOf at root (from z.union, z.discriminatedUnion, etc.)
              // See: https://github.com/anthropics/claude-code/issues/8014
              if (
                typeof convertedSchema === 'object' &&
                convertedSchema !== null &&
                'type' in convertedSchema &&
                convertedSchema.type === 'object'
              ) {
                outputSchema = convertedSchema as ToolOutput
              }
            }
            return {
              ...tool,
              description: await tool.prompt({
                getToolPermissionContext: async () => toolPermissionContext,
                tools,
                agents: [],
              }),
              inputSchema: zodToJsonSchema(tool.inputSchema) as ToolInput,
              outputSchema,
            }
          }),
        ),
      }
    },
  )

  server.setRequestHandler(
    CallToolRequestSchema,
    async ({ params: { name, arguments: args } }): Promise<CallToolResult> => {
      const toolPermissionContext = getEmptyToolPermissionContext()
      // TODO: Also re-expose any MCP tools
      const tools = getTools(toolPermissionContext)
      const tool = findToolByName(tools, name)
      if (!tool) {
        throw new Error(`Tool ${name} not found`)
      }

      // Assume MCP servers do not read messages separately from the tool
      // call arguments.
      const toolUseContext: ToolUseContext = {
        abortController: createAbortController(),
        options: {
          commands: MCP_COMMANDS,
          tools,
          mainLoopModel: getMainLoopModel(),
          thinkingConfig: { type: 'disabled' },
          mcpClients: [],
          mcpResources: {},
          isNonInteractiveSession: true,
          debug,
          verbose,
          agentDefinitions: { activeAgents: [], allAgents: [] },
        },
        getAppState: () => getDefaultAppState(),
        setAppState: () => {},
        messages: [],
        readFileState: readFileStateCache,
        setInProgressToolUseIDs: () => {},
        setResponseLength: () => {},
        updateFileHistoryState: () => {},
        updateAttributionState: () => {},
      }

      // TODO: validate input types with zod
      try {
        if (!tool.isEnabled()) {
          throw new Error(`Tool ${name} is not enabled`)
        }
        const validationResult = await tool.validateInput?.(
          (args as never) ?? {},
          toolUseContext,
        )
        if (validationResult && !validationResult.result) {
          throw new Error(
            `Tool ${name} input is invalid: ${validationResult.message}`,
          )
        }
        const finalResult = await tool.call(
          (args ?? {}) as never,
          toolUseContext,
          hasPermissionsToUseTool,
          createAssistantMessage({
            content: [],
          }),
        )

        return {
          content: [
            {
              type: 'text' as const,
              text:
                typeof finalResult === 'string'
                  ? finalResult
                  : jsonStringify(finalResult.data),
            },
          ],
        }
      } catch (error) {
        logError(error)

        const parts =
          error instanceof Error ? getErrorParts(error) : [String(error)]
        const errorText = parts.filter(Boolean).join('\n').trim() || 'Error'

        return {
          isError: true,
          content: [
            {
              type: 'text',
              text: errorText,
            },
          ],
        }
      }
    },
  )

  async function runServer() {
    const transport = new StdioServerTransport()
    await server.connect(transport)
  }

  return await runServer()
}