Getting Started
This page walks through connecting an LLM client to the UniBio Intelligence platform in under five minutes. It assumes you have an MCP-capable client installed (Claude Desktop, Cursor, the OpenAI agents SDK, or similar).
If you are not yet sure which component you need, see the overview first.
1. Pick the surfaces you want
| You want to ... | Connect to |
|---|---|
| Look up targets, drugs, trials, expression, structures | UbiMCP |
| Run antibody numbering, humanization, structure prediction, design | UbiTools MCP |
You can connect to both UbiMCP and UbiTools at the same time; they appear as two separate MCP servers in your client.
2. Endpoints
The hosted endpoints are:
| Service | URL |
|---|---|
| UbiMCP (data layer) | https://mcp.unibiointelligence.com/mcp |
| UbiTools (compute layer) | https://mcp.tools.unibiointelligence.com/mcp |
Both endpoints speak the Model Context Protocol over HTTP. UbiMCP also supports stdio when running locally; UbiTools requires authentication (see UbiTools authentication).
3. Connect a client
- Connect Claude Desktop
- Connect ChatGPT and the OpenAI Agents SDK
- Connect Cursor, VS Code, and other clients
4. Try it
Once connected, ask your LLM a question that requires data lookup. Examples:
- "What targets are most strongly associated with idiopathic pulmonary fibrosis in Open Targets?"
- "Pull the GTEx expression profile for TYK2 across tissues."
- "Find clinical trials for trastuzumab deruxtecan that are currently recruiting."
- "Number this antibody heavy chain using IMGT: EVQLVESGGGLVQPGG..."
The client will discover the relevant tool, call it, and ground its answer in the returned data.
5. Move on to the reference docs
- UbiMCP server catalog: the 18 databases available and what each one exposes.
- UbiTools MCP tool reference: the design and prediction tools surfaced to LLMs.