Understand AI memory
What is AI memory?
Understand the difference between a chat transcript, saved facts, and the context an assistant uses to answer your next question.
THE DRAY FIELD GUIDE
Less retelling.
More doing.
Find the old decision. Brief the next assistant. Keep control of what you share.
30 practical guides for work that spans AI tools.
Find a guidePICK A STARTING POINT
What are you working through?
THE LIBRARY
Find your next step.
Short guides. Useful templates.
No account needed to read or copy.
30 guides
Understand AI memory
AI memory vs. chat history
Decide when you need the original conversation and when a short memory record is enough, with a checklist for checking exact claims.
Understand AI memory
AI memory vs. a context window
Learn why a large context window does not replace saved memory, and how to choose the information a new chat needs first.
Understand AI memory
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
What does MCP do for AI memory?
A plain-language guide to MCP clients, servers, memory tools, and the difference between connecting an agent and capturing its work.
Understand AI memory
How to check the source of an AI memory
Trace an AI answer back to a conversation, distinguish a quote from a summary, and check whether a later decision changed the claim.
Move work between tools
Move project context from ChatGPT to Claude
Continue a ChatGPT project in Claude with a checked brief, source references, and a clear next task instead of a vague chat summary.
Move work between tools
Continue a Claude project in ChatGPT
Carry a Claude draft or analysis into ChatGPT while preserving corrections, references, and the difference between a draft and an approved decision.
Move work between tools
Bring project memory to a coding agent
Give a coding agent useful project decisions through a brief or MCP connection without assuming it has your repository, secrets, or chat history.
Move work between tools
A pasted brief or shared AI memory?
Choose between a manual project brief and shared memory based on how often you switch tools, how much context changes, and what you can share.
Move work between tools
What to check before sending an AI chat handoff
Review the destination, source scope, files, and next action before moving a conversation to another AI assistant.
Move work between tools
Write a project brief an AI can actually use
A reusable AI project brief with a clear goal, current decisions, constraints, evidence, and next task. Includes a copyable template.
Keep a project moving
Keep a decision log across AI conversations
Record what was decided, why, and what would change it so later chats do not reopen settled questions or mistake proposals for commitments.
Keep a project moving
Resume a coding task in a new AI chat
Leave a coding checkpoint that records the branch, changed files, test results, and next action without claiming unverified work is complete.
Keep a project moving
Organize research notes from several AI chats
Build a research record that separates original sources, assistant summaries, and unresolved claims before you write or make a decision.
Keep a project moving
Keep a brand voice across AI writing tools
Replace vague tone instructions with a short voice brief, approved examples, and constraints that travel between your AI writing tools.
Keep a project moving
Keep client context separate in AI work
Reduce accidental mixing of client details with explicit scope, approved source lists, and a checked brief before you use another AI tool.
Keep a project moving
Get a useful second opinion from another AI
Ask another assistant to test a specific assumption with the same evidence, rather than simply vote on the first assistant’s answer.
Find and fix a missing piece
Find a past AI conversation when you forgot its title
Search for an old AI chat using distinctive phrases, project details, and decision clues, then check the surrounding messages.
Find and fix a missing piece
Why is a chat missing from AI memory?
Check account, capture scope, sync, and retrieval separately when an expected conversation does not appear in AI memory.
Find and fix a missing piece
The assistant is connected but not using memory
Distinguish a connected account from a successful memory request, and test whether the assistant can retrieve a known source.
Find and fix a missing piece
Fix an outdated fact in AI memory
Correct an old deadline or project preference without losing the history of what changed, then test whether later answers use the correction.
Find and fix a missing piece
A chat handoff says the source is unavailable
Check source sync, destination access, and the prepared request before retrying an AI conversation handoff.
Find and fix a missing piece
Verify an AI memory connection end to end
Use a synthetic project to test capture, sync, tool access, and source-backed recall without exposing private conversation data.
Choose what gets shared
Share only the context an AI task needs
Choose between an excerpt, a project brief, and a whole-chat handoff based on the information the next assistant actually needs.
Choose what gets shared
Pausing AI recall is not the same as stopping capture
Understand the difference between collecting chats, using memory in answers, and making an explicit memory request in Dray.
Choose what gets shared
Deleting AI memory vs. revoking access
Work out whether you need to remove stored content, disconnect a client, stop new capture, or make a broader data request.
Choose what gets shared
Where does AI memory data go?
Follow the path from a browser conversation to memory processing and connected assistants, and identify the policy questions to ask.
Choose what gets shared
Before you enable file capture for AI memory
Check attachments, sharing permissions, destination access, and retention before allowing files to be copied into an AI memory service.
Choose what gets shared
Review permissions before connecting an AI tool
Check account identity, read and write authority, source scope, and revocation before approving a browser or MCP connection.