graphistry-mcp

Inspiration

We were inspired by the vision of a MCP that unifies human and AI agents into a seamless investigation and orchestration interface. At Graphistry, we’re always pushing the boundary on visual-first intelligence—so we challenged ourselves: what if we could talk to our data, query visual graphs, execute tools, and even speak to Louie.ai in a multimodal experience, all through a real-time control panel?

What it does

GPU-accelerated graph visualization via Graphistry Advanced pattern discovery and relationship analysis Streamable HTTP interface for resumable connections Support for various graph data formats (Pandas, NetworkX, edge lists) Interactive graph visualization and exploration Layout control for different visualization types Network investigation and anomaly detection capabilities

How we built it

We started by exposing graph + tool APIs.

Challenges we ran into

Managing Graphistry’s real-time rendering with live chat and voice threads simultaneously.

Deciding what actions the MCP should take autonomously versus when to pause for human approval.

Accomplishments that we're proud of

Built a real, AI + graph control system in a day

Created an experience where data analysis feels like commanding a sci-fi AI assistant.

Graphistry can be used together to saturate an investigation surface with minimal user friction.

What we learned

That mcp is amazingly fun

What's next for graphistry-mcp

We’re integrating this into Graphistry Hub. We’ll be extending it to include agent planning memory, user session journaling, and even a comic-strip style visual mode for past investigation replays. And yes—expect Bluetooth mic support and on-the-go investigation, all powered by GPU + AI.

Built With

• graphistry
• mcp
• python