Fix: Check for MCP Available Attributes

[ajshedivy]

Description

proposed fix for #3225

Solution

Since some servers do not provider resources, tools, etc., check if client methods are available with simple wrapper:

this.client.listResources()
this.client.listTools()
this.client.listPrompts()

  async checkClientAttribute(func: () => Promise<any>): Promise<any> {
    try {
      const result = await func();
      return result;
    } catch (e) {
      return null;
    }
  }

There is prob a better way to handle this, but it's a good hot fix 😁

Checklist

• [] The relevant docs, if any, have been updated or created

Screenshots

Testing

1. Install the filesystem server:

npm install -g @modelcontextprotocol/server-filesystem

2. Configure MCP in ~/.continue/config.json:

{
  "experimental": {
    "modelContextProtocolServer": {
      "transport": {
        "type": "stdio",
        "command": "npx",
        "args": [
          "-y",
          "@modelcontextprotocol/server-filesystem",
          "/path/to/directory"
        ]
      }
    }
  }
}

TODO

• handle duplicate Tools:
  Adding the @modelcontextprotocol/server-filesystem server produces the following warning with the default tools:

Warning: Encountered two children with the same key, `read_file`. 
Keys should be unique so that components maintain their identity across updates. 
Non-unique keys may cause children to be duplicated and/or omitted — the behavior is unsupported and could change in a future version.

[netlify]

✅ Deploy Preview for continuedev ready!

Name	Link
🔨 Latest commit	2861775
🔍 Latest deploy log	https://app.netlify.com/sites/continuedev/deploys/67577b658f455b0008b5b6c1
😎 Deploy Preview	https://deploy-preview-3232--continuedev.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[ajshedivy]

To handle duplicate tools, an easy fix would just be to encode the BuiltinTools using some sort of prefix:

export enum BuiltInToolNames {
  ReadFile = "builtin_read_file",
  ReadCurrentlyOpenFile = "builtin_read_currently_open_file",
  CreateNewFile = "builtin_create_new_file",
  RunTerminalCommand = "builtin_run_terminal_command",
  ViewSubdirectory = "builtin_view_subdirectory",
  ViewRepoMap = "builtin_view_repo_map",
  ExactSearch = "builtin_exact_search",
  SearchWeb = "builtin_search_web",
  ViewDiff = "builtin_view_diff",
}

[ajshedivy]

@sestinj @Patrick-Erichsen This is ready for review!

[SimonB97]

I tried to get this to work on my end, could you confirm my config.json structure please? I put "experimental" conf at the root (and at the end).

{
  "models": [
    {
      "title": "Claude 3.5 Sonnet (Free Trial)",
      "provider": "free-trial",
      "model": "claude-3-5-sonnet-latest",
      "systemMessage": "You are an expert software developer. You give helpful and concise responses."
    },
    {
      "title": "GPT-4o (Free Trial)",
      "provider": "free-trial",
      "model": "gpt-4o",
      "systemMessage": "You are an expert software developer. You give helpful and concise responses."
    },
    {
      "title": "Llama3.1 70b (Free Trial)",
      "provider": "free-trial",
      "model": "llama3.1-70b",
      "systemMessage": "You are an expert software developer. You give helpful and concise responses."
    },
    {
      "title": "Codestral (Free Trial)",
      "provider": "free-trial",
      "model": "codestral-latest",
      "systemMessage": "You are an expert software developer. You give helpful and concise responses."
    },
    {
      "model": "claude-3-5-sonnet-latest",
      "provider": "anthropic",
      "apiKey": "",
      "title": "Claude 3.5 Sonnet"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Tab Autocomplete",
    "provider": "free-trial",
    "model": "codestral-latest"
  },
  "customCommands": [
    {
      "name": "test",
      "prompt": "{{{ input }}}\n\nWrite a comprehensive set of unit tests for the selected code. It should setup, run tests that check for correctness including important edge cases, and teardown. Ensure that the tests are complete and sophisticated. Give the tests just as chat output, don't edit any file.",
      "description": "Write unit tests for highlighted code"
    }
  ],
  "contextProviders": [
    {
      "name": "code",
      "params": {}
    },
    {
      "name": "docs",
      "params": {}
    },
    {
      "name": "diff",
      "params": {}
    },
    {
      "name": "terminal",
      "params": {}
    },
    {
      "name": "problems",
      "params": {}
    },
    {
      "name": "folder",
      "params": {}
    },
    {
      "name": "codebase",
      "params": {}
    }
  ],
  "slashCommands": [
    {
      "name": "share",
      "description": "Export the current chat session to markdown"
    },
    {
      "name": "cmd",
      "description": "Generate a shell command"
    },
    {
      "name": "commit",
      "description": "Generate a git commit message"
    }
  ],
  "experimental": {
    "modelContextProtocolServer": {
      "transport": {
        "type": "stdio",
        "command": "npx",
        "args": [
          "-y",
          "@modelcontextprotocol/server-filesystem",
          "~/Desktop"
        ]
      }
    }
  },
  "embeddingsProvider": {
    "provider": "free-trial"
  },
  "reranker": {
    "name": "free-trial"
  }
}

[Patrick-Erichsen]

@adamszewe thanks for opening this!

I poked around the TS interface for client a bit and I think we can use client.getServerCapabilities in the modifyConfig to do this a bit cleaner.

https://github.com/search?q=repo%3Amodelcontextprotocol%2Ftypescript-sdk%20getServerCapabilities&type=code

--
150 | * After initialization has completed, this will be populated with the server's reported capabilities.
151 | */
152 | getServerCapabilities(): ServerCapabilities \| undefined {
153 | return this._serverCapabilities;
154 | }


// Server supports resources and tools, but not prompts
--
214 | expect(client.getServerCapabilities()).toEqual({
215 | resources: {},
216 | tools: {},
217 | });

I think our logic could then look something like this:

    const capabilities = this.client.getServerCapabilities()

    if (capabilities?.resources) {
      const resources = this.client.listResources()
      // ...
    }

+1 to the implementation for BuiltInToolNames, I think the prefix is a good call.

[sestinj]

@ajshedivy I think the right way to do this is probably with the client.getServerCapabilities() method on the MCP client. Any interest in attempting that?

Edit: totally missed that Patrick caught this above, ignore me!

[ajshedivy]

@adamszewe thanks for opening this!

I poked around the TS interface for client a bit and I think we can use client.getServerCapabilities in the modifyConfig to do this a bit cleaner.

https://github.com/search?q=repo%3Amodelcontextprotocol%2Ftypescript-sdk%20getServerCapabilities&type=code

--
150 | * After initialization has completed, this will be populated with the server's reported capabilities.
151 | */
152 | getServerCapabilities(): ServerCapabilities \| undefined {
153 | return this._serverCapabilities;
154 | }


// Server supports resources and tools, but not prompts
--
214 | expect(client.getServerCapabilities()).toEqual({
215 | resources: {},
216 | tools: {},
217 | });

I think our logic could then look something like this:

    const capabilities = this.client.getServerCapabilities()

    if (capabilities?.resources) {
      const resources = this.client.listResources()
      // ...
    }

+1 to the implementation for BuiltInToolNames, I think the prefix is a good call.

This would be much cleaner 😁

@adamszewe thanks for opening this!

I poked around the TS interface for client a bit and I think we can use client.getServerCapabilities in the modifyConfig to do this a bit cleaner.

https://github.com/search?q=repo%3Amodelcontextprotocol%2Ftypescript-sdk%20getServerCapabilities&type=code

--
150 | * After initialization has completed, this will be populated with the server's reported capabilities.
151 | */
152 | getServerCapabilities(): ServerCapabilities \| undefined {
153 | return this._serverCapabilities;
154 | }


// Server supports resources and tools, but not prompts
--
214 | expect(client.getServerCapabilities()).toEqual({
215 | resources: {},
216 | tools: {},
217 | });

I think our logic could then look something like this:

    const capabilities = this.client.getServerCapabilities()

    if (capabilities?.resources) {
      const resources = this.client.listResources()
      // ...
    }

+1 to the implementation for BuiltInToolNames, I think the prefix is a good call.

@Patrick-Erichsen Good find! much better solution 😁

[YanxingLiu]

I tried to get this to work on my end, could you confirm my config.json structure please? I put "experimental" conf at the root (and at the end).

{
  "models": [
    {
      "title": "Claude 3.5 Sonnet (Free Trial)",
      "provider": "free-trial",
      "model": "claude-3-5-sonnet-latest",
      "systemMessage": "You are an expert software developer. You give helpful and concise responses."
    },
    {
      "title": "GPT-4o (Free Trial)",
      "provider": "free-trial",
      "model": "gpt-4o",
      "systemMessage": "You are an expert software developer. You give helpful and concise responses."
    },
    {
      "title": "Llama3.1 70b (Free Trial)",
      "provider": "free-trial",
      "model": "llama3.1-70b",
      "systemMessage": "You are an expert software developer. You give helpful and concise responses."
    },
    {
      "title": "Codestral (Free Trial)",
      "provider": "free-trial",
      "model": "codestral-latest",
      "systemMessage": "You are an expert software developer. You give helpful and concise responses."
    },
    {
      "model": "claude-3-5-sonnet-latest",
      "provider": "anthropic",
      "apiKey": "",
      "title": "Claude 3.5 Sonnet"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Tab Autocomplete",
    "provider": "free-trial",
    "model": "codestral-latest"
  },
  "customCommands": [
    {
      "name": "test",
      "prompt": "{{{ input }}}\n\nWrite a comprehensive set of unit tests for the selected code. It should setup, run tests that check for correctness including important edge cases, and teardown. Ensure that the tests are complete and sophisticated. Give the tests just as chat output, don't edit any file.",
      "description": "Write unit tests for highlighted code"
    }
  ],
  "contextProviders": [
    {
      "name": "code",
      "params": {}
    },
    {
      "name": "docs",
      "params": {}
    },
    {
      "name": "diff",
      "params": {}
    },
    {
      "name": "terminal",
      "params": {}
    },
    {
      "name": "problems",
      "params": {}
    },
    {
      "name": "folder",
      "params": {}
    },
    {
      "name": "codebase",
      "params": {}
    }
  ],
  "slashCommands": [
    {
      "name": "share",
      "description": "Export the current chat session to markdown"
    },
    {
      "name": "cmd",
      "description": "Generate a shell command"
    },
    {
      "name": "commit",
      "description": "Generate a git commit message"
    }
  ],
  "experimental": {
    "modelContextProtocolServer": {
      "transport": {
        "type": "stdio",
        "command": "npx",
        "args": [
          "-y",
          "@modelcontextprotocol/server-filesystem",
          "~/Desktop"
        ]
      }
    }
  },
  "embeddingsProvider": {
    "provider": "free-trial"
  },
  "reranker": {
    "name": "free-trial"
  }
}

Hello, have you solved the problem. I encountered the same issue that I have added the modelContextProtocolServers, but I cannnot see the mcp after I typed @.

[SimonB97]

Tried last time a few days ago (maybe they have updated it since) but sadly still didn't work for me

[ajshedivy]

@SimonB97 @YanxingLiu the config has actually changed in the recent pre-release:
change modelContextProtocolServer -> modelContextProtocolServers (severs in now plural). I opened a PR with some potential fixes for multiple servers: #3462

[YanxingLiu]

Contributor

Thanks for your work, I have tried the setting mentioned in #3462. But it still failed. The detail is shown in that PR. I would like to ask if you know why this is. Thanks a lot for your help.