MCP Server: Ollama Deep Researcher

What is Ollama Deep Researcher?

Ollama Deep Researcher is a Model Context Protocol (MCP) server adaptation of LangChain Ollama Deep Researcher. It provides deep research capabilities as MCP tools, allowing AI assistants to perform in-depth, iterative research on topics using local LLMs hosted by Ollama and web search via Tavily or Perplexity.

How to use Ollama Deep Researcher?

Clone the repository, install Node.js and Python dependencies (via npm, uv or pip), build the TypeScript code, and pull an Ollama model. Add the server to your MCP client’s configuration (e.g., Claude Desktop or Cline) using either the standard or Docker installation. Use the research tool with a topic, optionally configure parameters with the configure tool, and access results as MCP resources.

Key features of Ollama Deep Researcher

• Iterative research process with multiple cycles of search, summarization, and reflection
• Uses local LLMs via Ollama (e.g., deepseek-r1:1.5b)
• Web search via Tavily or Perplexity API
• Configurable number of research iterations (1–5)
• Outputs a markdown summary with source citations
• Supports Docker deployment
• Integrates with LangSmith for tracing and monitoring

Use cases of Ollama Deep Researcher

• AI assistants performing deep, multi‑step research on a given topic
• Researching complex subjects that require iterative refinement
• Generating comprehensive summaries with citations for knowledge work
• Providing persistent, reusable research resources within MCP clients

FAQ from Ollama Deep Researcher

What prerequisites are required to run Ollama Deep Researcher?

You need Node.js, Python 3.10+, an Ollama model, and API keys for Tavily and Perplexity. At least 8GB of RAM is recommended for larger models.

Which search APIs does Ollama Deep Researcher support?

The server supports Tavily and Perplexity API, configured via the searchApi argument in the configure tool.

Can I run the server with Docker?

Yes, a Docker installation is provided. You build the container, set API keys in a .env file, and run helper scripts like run-docker.sh start.

How are research results stored?

Completed research is stored as MCP resources accessible via research://{topic} URIs. They appear in the MCP client’s resource panel with timestamps and descriptions.

What tracing and monitoring is available?

Ollama Deep Researcher integrates with LangSmith to trace all LLM interactions, web searches, and research workflow steps for performance monitoring and debugging.