ollama handoff for VS Code

io.github.michael-whitecapdata/ollama-handoff

Offload cheap work from your AI agent to a local Ollama model, at zero cloud cost.

client:VS Code transport:stdio runtime:pypi

Install ollama handoff in VS Code

.vscode/mcp.json

{
  "inputs": [
    {
      "type": "promptString",
      "id": "ollama-url",
      "description": "OLLAMA_URL",
      "password": true
    },
    {
      "type": "promptString",
      "id": "ollama-default-model",
      "description": "OLLAMA_DEFAULT_MODEL",
      "password": true
    },
    {
      "type": "promptString",
      "id": "ollama-num-ctx",
      "description": "OLLAMA_NUM_CTX",
      "password": true
    },
    {
      "type": "promptString",
      "id": "ollama-keep-alive",
      "description": "OLLAMA_KEEP_ALIVE",
      "password": true
    },
    {
      "type": "promptString",
      "id": "ollama-timeout-s",
      "description": "OLLAMA_TIMEOUT_S",
      "password": true
    }
  ],
  "servers": {
    "ollama-handoff": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "ollama-handoff"
      ],
      "env": {
        "OLLAMA_URL": "${input:ollama-url}",
        "OLLAMA_DEFAULT_MODEL": "${input:ollama-default-model}",
        "OLLAMA_NUM_CTX": "${input:ollama-num-ctx}",
        "OLLAMA_KEEP_ALIVE": "${input:ollama-keep-alive}",
        "OLLAMA_TIMEOUT_S": "${input:ollama-timeout-s}"
      }
    }
  }
}

VS Code prompts for each input the first time the server starts, so no secret is written to the file. VS Code docs

ollama handoff in other clients

ollama handoff MCP server