Pollinations Think

Provides advanced MCP thinking and web-search capabilities using Pollinations AI with DeepSeek reasoning models.

javascript

0

GitHub Stars

javascript

Language

6 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "palolxx-pollinations-think-mcp": {
      "command": "node",
      "args": [
        "/path/to/pollinations-think-mcp/index.js"
      ]
    }
  }
}

You can run a dedicated MCP server that provides advanced thinking and analysis by interfacing with the Pollinations AI API using DeepSeek reasoning models. This server supports multi-cycle thinking, robust error handling, and health checks, making it suitable for scenarios that require strategic analysis, real-time web search, and flexible model choices. It is designed to be deployed as part of your MCP ecosystem and can be wired into your MCP client configuration to perform high-level reasoning and information gathering.

How to use

To use this MCP server, connect your MCP client to the local or remote stdio endpoint described in your configuration. The server exposes tools for advanced thinking and web search, enabling you to perform multi-cycle reasoning, detect contradictions, and synthesize results across cycles. When you issue a think command, you can specify the topic or question, choose a model, set a seed for reproducibility, and control the number of thinking cycles. Use the search capability to pull in real-time information, then continue receiving the next part of a large thinking response with the appropriate continuation tool. If a response is large, it will be split and you can request subsequent parts as needed.

Using the available tools

Think drives deep, multi-cycle reasoning using a reasoning model. You provide the text prompt and optionally select a model, seed, and maximum cycles. The server then returns a structured analytical result with an executive summary, risk assessment, and next steps.

Search performs real-time web lookups using a web-search model to retrieve current information and sources. This helps ground decisions with up-to-date data.

Continue_thinking lets you fetch the next portion of a large thinking response when the initial output is cut due to length limitations.

Installation prerequisites and startup

You will need a Node.js runtime to run the MCP server as a local stdio process.

Environment and deployment notes

The server is designed to be deployed in environments compatible with MCP clients. No additional environment variables are required to run, since it uses public Pollinations.ai endpoints.

Available tools

think

Advanced strategic thinking and analysis using an openai-reasoning model. You provide the topic, optional model, seed, and maxCycles to control multi-cycle thinking.

search

Real-time web search using a web-search model to fetch current information and sources.

continue_thinking

Continue receiving the next portion of a large thinking response when the initial output is split due to length constraints.