How shared memory between AI assistants works
See what a shared memory connection can provide, what it cannot transfer, and how to test access with a harmless project detail.
Understand AI memory 2 min read
What does MCP do for AI memory?
MCP, the Model Context Protocol, is a standard for connecting AI applications to external tools and data. A memory service can expose tools through MCP so an authorized agent can search or read stored context.
The connection is not the memory
The AI application is the client. The service offering data or tools is the server. MCP describes how they communicate. It does not decide which facts should be remembered, how long data is kept, or whether a retrieved result is correct.
An application supporting MCP is only one prerequisite. Its configuration, permissions, and available features still affect whether a particular service can be used. Follow the instructions for your client rather than pasting a configuration written for another tool.
Reading context does not imply uploading a workspace
A tool that searches memory needs access to stored records. Capturing a terminal session or reading a local repository is a separate capability. Do not infer that one happens because the other is connected.
Test both authorization and actual tool use. An account can appear connected while a tool request fails. The useful evidence is a successful request with a result you can trace, not just a status badge.
TAKE THIS WITH YOU
Questions to ask before adding a server
An illustrative template. Replace the details with material you are allowed to share.
Which tools will this server expose? What account and data can they access? Can they only read, or can they change data? Which actions need my approval? How do I revoke access? What harmless request will prove the setup works?
USING THIS WITH DRAY
Keep the context. Check the source.
Dray supports stored-context access through MCP. Use the current client-specific instructions for setup; the web extension and experimental agent channels are separate parts of the beta.
Dray is in beta. Available sources and connections depend on your setup. Read the data notes before granting access.
Sources & next reads
Product details are grounded in the guides below. Example projects and prompts are illustrative, not customer results.