Context+

Install

npx skillfoxx add mcp/context

claude mcp add --transport stdio contextplus -- npx -y contextplus

{
  "mcpServers": {
    "contextplus": {
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

{
  "mcpServers": {
    "contextplus": {
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

code --add-mcp '{"name":"contextplus","type":"stdio","command":"npx","args":["-y","contextplus"]}'

{
  "servers": {
    "contextplus": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

codex mcp add contextplus -- npx -y contextplus

[mcp_servers.contextplus]
command = "npx"
args = ["-y", "contextplus"]

{
  "mcpServers": {
    "contextplus": {
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

{
  "mcpServers": {
    "contextplus": {
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

{
  "mcpServers": {
    "contextplus": {
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

{
  "mcpServers": {
    "contextplus": {
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

{
  "mcp": {
    "contextplus": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "contextplus"
      ]
    }
  }
}

{
  "context_servers": {
    "contextplus": {
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

{
  "mcpServers": {
    "contextplus": {
      "command": "npx",
      "args": [
        "-y",
        "contextplus"
      ]
    }
  }
}

{
  "mcpServers": {
    "contextplus": {
      "command": "npx",
      "args": ["-y", "contextplus"],
      "env": {
        "OLLAMA_EMBED_MODEL": "nomic-embed-text"
      }
    }
  }
}

This is third-party code. Review the repository files before installing.

What it does

Context+ indexes a project and gives the agent tools for a large codebase. Discovery tools build a structural AST tree, show signatures without function bodies, search by meaning through embeddings and group related files with spectral clustering. Analysis tools trace where a symbol is used and run linters and compilers to find dead code and type errors in TypeScript, Python, Rust and Go. Writing code goes only through the propose_commit tool, which validates edits and creates a restore point before saving, with a list of restore points and undo. A separate module keeps memory as a graph of nodes and relations with embeddings and neighbor traversal. Embeddings run on a local Ollama by default or through an OpenAI-compatible API.

Who it is for. For developers working on large projects who want the agent to navigate by meaning, not just by text.

Example request

Find by meaning where the project handles file uploads and show every place that function is used

Limitations

Embeddings need a running local Ollama or a key to an OpenAI-compatible provider. Cloud providers such as OpenAI and Gemini require access to their networks, a local Ollama works offline. Indexing a large repository takes time. Code writes go only through propose_commit.

How to disable. Remove the contextplus block from your client's MCP configuration and restart it.

MCP

Transport

stdio

Authentication

not required

Security check

• Reads project files and runs linters and compilers for analysis
• Changes code files through propose_commit and reaches the network for embeddings with a cloud provider

README in short

The README describes Context+ as an MCP server that combines RAG, Tree-sitter AST, spectral clustering and Obsidian-style linking to make a large codebase navigable. Tools are grouped into discovery, analysis, code ops, version control through restore points and a graph-based memory. Installation is via npx or bunx by adding a block to the client's MCP config, with an init subcommand to generate the config for Claude Code, Cursor, VS Code, Windsurf and OpenCode. Embeddings support two backends: a local Ollama by default and an OpenAI-compatible API. MIT licensed.

Related