Ollama Deep Researcher

Ollama Deep Researcher is an MCP server that facilitates autonomous topic research by integrating local large language models through Ollama with web search APIs like Tavily, Perplexity, and Exa. Designed for researchers, developers, and analysts who require iterative web querying and synthesis without sending model prompts to proprietary cloud LLMs, the server manages the multi-step research workflow locally over standard input and output. The system coordinates query expansion, search execution, source extraction, and document synthesis while providing tools to track progress and adjust runtime parameters. Users can specify target topics, set maximum research iteration loops, select custom language models, and choose specific search backends depending on the semantic depth required. Operating as a Desktop Extension compatible local server, it exposes research findings directly as accessible MCP resources to connected client applications. It also features built-in process management, such as thirty-minute subprocess timeouts and comprehensive error handling directed to standard error streams, ensuring stable local execution during extensive literature reviews or market intelligence workflows.

Category: AI & LLM Tooling

Tags: llm, ollama, perplexity, research, tavily

Visit Ollama Deep Researcher

How to install and configure Ollama Deep Researcher

1. Clone the repository and install project dependencies: sh git clone https://github.com/Cam10001110101/mcp-server-ollama-deep-researcher cd mcp-server-ollama-deep-researcher npm install 2. Install the required Python assistant dependencies: sh cd src/assistant pip install -r requirements.txt cd ../.. 3. Export your search engine API keys and optional Ollama endpoint: sh export TAVILY_API_KEY=your_tavily_key export PERPLEXITY_API_KEY=your_perplexity_key export EXA_API_KEY=your_exa_key export OLLAMA_BASE_URL=http://localhost:11434 4. Build the TypeScript MCP server: sh npm run build 5. Register the server in your MCP client configuration (such as Claude Desktop or a DXT host) using stdio transport pointing to node dist/index.js or via op run if resolving secrets using 1Password.

What you can do with Ollama Deep Researcher

• Executing iterative multi-loop web research on complex technical topics using local Ollama models alongside Tavily or Perplexity search APIs. - Extracting and summarizing up-to-date market intelligence reports and academic discoveries through semantic queries executed via the Exa neural search backend. - Monitoring and tracking the live execution status of background research subprocesses directly through standard MCP client interfaces and tools. - Configuring custom LLM synthesis models and iteration depth thresholds on a per-investigation basis to balance research thoroughness and local processing time.

Key facts

• https://github.com/Cam10001110101/mcp-server-ollama-deep-researcher
• AI & LLM Tooling, Web Search & Research
• llm, ollama, perplexity, research, tavily

Part of MCP Servers

What can Ollama Deep Researcher do?

Ollama Deep Researcher conducts in-depth topic investigations by combining local language models hosted in Ollama with external web search APIs including Tavily, Perplexity, and Exa. It runs configurable research loops, synthesizes retrieved web information, outputs structured summaries, and exposes findings directly to MCP clients as inspectable resources.

How do I install Ollama Deep Researcher?

You install the server by cloning its repository from GitHub, running npm install for TypeScript dependencies, and running pip install on requirements.txt inside the assistant directory. Next, set your preferred search API keys such as TAVILY_API_KEY or PERPLEXITY_API_KEY, build the project with npm run build, and register the output node script with your MCP host.

Which search APIs does Ollama Deep Researcher support?

The server supports Tavily for fast web search with raw content extraction, Perplexity for AI-generated search summaries with citations, and Exa for semantic neural search highlighting relevant passages. You can select your preferred provider by setting the appropriate environment API key and configuring the searchApi tool parameter.

Which MCP clients work with Ollama Deep Researcher?

Ollama Deep Researcher works with any client supporting the Model Context Protocol over standard input and output transport, including Claude Desktop and hosts supporting the Desktop Extension ecosystem. It also supports launch configurations managed through 1Password CLI tooling or Docker Compose environments.