Open-source MCP knowledge infrastructure
Build knowledge
your tools can
actually use.
Turn repositories, papers, documentation, and notebooks into a searchable knowledge base for agents, applications, and research teams.
Repositories
Code, docs, READMEs, and notebooks
Research papers
PDF extraction with external OCR support
Semantic retrieval
Search and retrieve precise source passages
MCP + HTTP
Serve agents, editors, and applications
Quick start
# Install
pip install nancy-brain
# Initialize a project
nancy-brain init my-knowledge-base
cd my-knowledge-base
# Edit config/repositories.yml to add your repos
nancy-brain build
# Search your knowledge base
nancy-brain search "machine learning algorithms"
Choose your route
Create an index
Install the package, configure source repositories, and build a local knowledge base.
Follow the quick start →Give agents access
Connect VS Code, Claude Desktop, Gemini, or another MCP-capable client.
Browse integrations →Run it persistently
Deploy the MCP and HTTP service with authentication, monitoring, and an admin UI.
Read the deployment guide →Four ways in
| Interface | Best for |
|---|---|
| MCP | Remote and local agent clients using Streamable HTTP or stdio |
| HTTP API | Search, retrieve, tree, weighting, status, and custom applications |
| CLI | Building, inspecting, testing, and querying indexes |
| Admin UI | Operational inspection and controlled rebuilds |
Start with the installation guide for prerequisites, use the research workflow for an end-to-end example, or go directly to the CLI reference.
Nancy Brain is named for Nancy Grace Roman, the astronomer who helped turn ambitious space observatories into working scientific infrastructure.