Town Library
Town Library is the easiest way to index your private, unstructured data and make it accessible to AI agents.
Connect your email, documents, calendar, and chat — Library handles the rest. Your data becomes instantly searchable by meaning, not just keywords.
Two Ways to Use Your Data
Once your data is indexed, you can use it in two ways:
1. Deploy a Research Agent
Create a fully-featured research agent with one click. Your agent comes with:
- Intelligent research — Multi-step reasoning with parallel subagents
- Beautiful chat interface — Modern web UI with markdown rendering, sources panel, and todo tracking
- All the tools — Web search, code execution, browser automation, and your Library data
- Adjustable depth — Fast answers, balanced research, or deep investigation
Learn more about Research Agents →
2. Connect via MCP
Expose your Library collections as MCP (Model Context Protocol) servers. Connect to:
- Claude — Desktop or web
- ChatGPT — Any interface that supports tools
- Cursor — IDE-integrated agent access
- Any MCP-compatible system — Your agents, third-party tools, custom integrations
Learn more about MCP Integration →
How It Works
| Step | What Happens |
|---|---|
| 1. Connect | Link your data sources via OAuth (Gmail, Drive, Calendar, Slack) or upload files directly |
| 2. Index | Library processes your data into semantic vectors that capture meaning |
| 3. Search | Agents query by concepts — “Q4 budget discussions” finds relevant content even without exact keywords |
| 4. Use | Deploy a research agent or connect external AI tools via MCP |
Quick Reference
| Term | Meaning |
|---|---|
| Collection | A set of connected data sources that agents can search |
| Data Source | A connected service (Gmail, Drive, Slack) or uploaded files |
| Search Index | The processed, searchable representation of your data |
| MCP | Model Context Protocol — connect Library to external agents |
| Research Agent | A deployed Town agent with full UI and research capabilities |
Start Here
- Getting Started — Create your first collection and index your data
- Collections — Organize data into searchable sets
- Data Sources — Connect Gmail, Drive, Calendar, Slack, or upload files
- Agents — Deploy a research agent
- MCP Integration — Connect Claude, Cursor, and other AI tools
Technical deep dive? Read How It Works for details on the indexing pipeline, knowledge graph, entity resolution, and search architecture.
Last updated on