AI connectivity: connect the AI your teams use to your R&D data
Teams already work in AI clients like Claude and ChatGPT, on models their company has approved, and across systems like SharePoint. The R&D data that would make those tools useful often sits out of reach, or can only be reached through a shared account that ignores who is asking.
Connect the AI your teams use to your R&D data is a three-page overview of the three paths between AI and Uncountable. The MCP Connector brings outside AI clients into Uncountable. BYOM (bring your own model) sets the models that Bodie, the Uncountable AI agent, runs on. Bodie Plugins bring your other systems into Bodie. Two worked examples follow a formulation scientist asking Claude about a project and a project lead asking Bodie a question that needs both SharePoint and Uncountable.
Inside this solution brief:
- A guided setup wizard for Claude, ChatGPT, Copilot Studio, and Gemini Enterprise
- More than 90 MCP tools that read and update experiments, measurements, projects, and notebooks as the signed-in user
- Admin control to offer only read-only tools or turn off single tools
- Bodie on your own OpenAI, Anthropic, or Google API key, on Azure API Management, or on AWS Bedrock inference profiles
- Uncountable-managed model connections beside your own, which admins add, reorder, and test for each account
- Bodie Plugins from the built-in catalog or any MCP server, such as SharePoint signed in with Microsoft Entra
- Per-user tokens, with no shared service account
- A side-by-side comparison of what each path connects, who sets it up, and whose access it uses
See how your teams can bring AI to R&D data without widening anyone's access.
FAQs
It is a three-page overview of the three ways Uncountable connects to AI: the MCP Connector for outside AI clients, BYOM for the models Bodie runs on, and Bodie Plugins for other systems. It includes a comparison table and two worked examples.
An admin sets up Claude, ChatGPT, Copilot Studio, or Gemini Enterprise in a step-by-step OAuth wizard. Each user then signs in, and more than 90 tools read and update experiments, measurements, projects, and notebooks as that user.
BYOM (bring your own model) runs Bodie, the Uncountable AI agent, on your own OpenAI, Anthropic, or Google API key, on OpenAI models behind your Azure API Management, or on your AWS Bedrock inference profiles. Uncountable-managed model connections are also available, and admins add, reorder, and test the connectors for each account.
Bodie Plugins connect Bodie to the MCP servers of your other systems, such as SharePoint. An admin adds a plugin from the built-in catalog or by its address, each user signs in with their own account, and Bodie calls the plugin tools in chat with that user's token.
No. Each MCP call runs as the signed-in user, and each Bodie Plugin uses the user's own token, so a user's connection gives no other user access. Bodie runs in your tenant, formulation data never trains foundation models, and Uncountable holds SOC 2 Type II and ISO 27001.
