Skip to main content

Jama Connect Interchange Help

Frequently Asked Questions

Question

Answer

What is MCP?

Model Context Protocol (MCP) is an open protocol that enables seamless integration between LLM applications and external data sources and tools. See https://modelcontextprotocol.io/specification/2025-06-18 for more information.

Will MCP have access to Jama Connect AdvisorTM features?

No. MCP is used with external tools and isn’t exposed to the Jama Connect Advisor capabilities.

Do I need an additional server to operate the AI software? If so, what type of system specs might I need?

We are not planning to host or deploy an LLM for MCP to use. You need to host an LLM and work with the selected vendor to determine the required specifications.

What are the software applications I need to install for my Jama Connect server to use Jama Connect AI capabilities?

None, other than the LLM you choose to use. If you choose to use an MCP client, you must install that, but it is not a requirement to use MCP.

Can I install my LLM on my Jama Connect server?

No. We recommend installing LLM on a separate server from Jama Connect to avoid significant performance issues.

What does the MCP architecture look like?

MCP in our implementation is not a separate running server. Instead, it is exposed as an endpoint within the Jama Connect application at /rest/mcp. This endpoint receives MCP RPC messages and processes them according to the MCP specification listed in the first question.

Because of this architecture, MCP traffic follows the same transport path and infrastructure as standard REST API requests:

  • Requests arrive over HTTPS.

    • They pass through the same network path, gateways, and bulkheads.

    • They are processed by the same application services.

  • The difference between REST API and MCP is primarily the message structure.

    • REST API uses resource-oriented HTTP requests, such as GET, POST, and PUT.

    • MCP sends RPC-style messages defined by the MCP protocol.

  • However, both ultimately travel over standard HTTP(S) and are handled by the same infrastructure.

Could MCP impact performance or “take down” a tenant, and how does it compare to REST API?

No. MCP isn’t expected to create meaningful performance risk. In our implementation, MCP runs through the same infrastructure and controls as REST API, and its expected traffic volume is significantly lower than typical REST integrations. MCP is not making REST API calls or consuming REST API bandwidth.

Key points:

  • Same infrastructure as REST — MCP requests are received through an endpoint in the Jama Connect application (/mcp/mcp-core) and travel over standard HTTPS, using the same network path, bulkheads, and infrastructure protections as REST API.

  • Significantly lower request volume — MCP workflows typically generate tens of requests, while many REST API integrations generate thousands in the same timeframe.

  • Bulkhead protections already proven — Jama Connect infrastructure is designed to handle millions of REST API calls, so MCP traffic represents only a small incremental load.

    • Cloud bulkheads are currently handling between 8-9.2M REST calls per day.

    • Self-hosted capacity largely depends on your infrastructure. Contact Jama Software Support for more information and guidance in this area.

  • LLM usage cost limits runaway scenarios — If an MCP-driven workflow were to loop, it typically generates only 1–2 calls per second, and the cost of LLM usage limits how long such a loop might run.

MCP uses the same infrastructure protections as REST API, but generates far fewer requests. Under normal use, there is no realistic scenario where MCP degrades tenant performance or takes down a tenant. 

MCP operations are not designed to return extremely large datasets.

In this section: