Liana MCP Server

Overview

Liana‑MCP delivers a natural‑language interface for single‑cell RNA sequencing (scRNA‑Seq) analysis by exposing the rich functionality of the Liana framework through the Model Context Protocol. It solves a common bottleneck in bioinformatics pipelines: the need for domain experts to translate complex R or Python workflows into executable commands. By turning Liana’s analysis routines—data ingestion, cell‑cell communication inference, and visualisation—into MCP tools, developers can embed sophisticated scRNA‑Seq processing directly into conversational AI assistants, chatbots or agent systems without requiring users to write code.

The server provides three core capabilities that are immediately useful for researchers and developers. First, an IO module handles reading and writing of scRNA‑Seq datasets in standard formats (e.g., h5ad, loom), enabling seamless data flow between the client and the analysis engine. Second, it offers a cell‑cell communication method that implements Liana’s inference algorithms, allowing users to query intercellular signalling networks through plain language prompts. Third, a plotting module exposes several visualization primitives—circle plots and dotplots—that can be generated on demand and returned as image payloads. Together, these features give developers a full analytical stack that can be invoked from any MCP‑compatible client such as Cherry Studio, Cline, or Agno.

In practice, Liana‑MCP is ideal for building AI assistants that guide biologists through exploratory data analysis. For example, a user could ask the assistant to “identify ligand‑receptor interactions between T cells and macrophages in my dataset,” and the server would execute Liana’s inference, return a ranked list of interactions, and optionally generate a dotplot visualizing the top signals. The same workflow can be embedded in an automated research pipeline where agents iterate over multiple samples, aggregate results, and generate publication‑ready figures—all without manual scripting.

Integration is straightforward: the MCP client declares a server entry with either a local command or a remote URL, and the assistant can call the defined tools using natural language. Because MCP abstracts away transport details, Liana‑MCP can run on a local workstation or be exposed as a remote service over HTTP, making it flexible for both individual researchers and institutional deployments. Its unique advantage lies in coupling Liana’s proven cell‑cell communication framework with the conversational flexibility of MCP, thereby democratizing advanced single‑cell analysis for non‑programmers while still offering full programmability for developers.