Consulting memory is not merely a list of personal preferences. It includes decisions, assumptions, open questions, promised follow-ups, source documents, definitions and the current state of a deliverable. Mixing those records across clients creates quality, confidentiality and billing risks.
A project-scoped memory layer can make continuity more deliberate: recall the project brief at the start, search client-authorized files, save established decisions as work progresses, and compact the brief when the engagement changes. The workflow must remain subordinate to the contract, data-handling rules and the consultant's professional judgment.
How should a consultant separate projects?
Use one Central Brain project per client, engagement or access boundary rather than one shared project for every engagement.
Choose a dedicated source folder and prepared-output location for each project. Give folders clear names that do not expose unnecessary confidential information. Keep engagement-level access, retention, backup and deletion instructions with the project operating record. Do not place a file in the project merely because it might be useful; confirm that the engagement permits its use.
Project-specific connections should point only to the intended project. If a supported AI tool is used for several clients, verify the active project before retrieval or checkpointing. A technical boundary helps limit scope, but the person remains responsible for choosing the correct project and account.
What should be saved as durable consulting memory?
Save established decisions, verified facts, open questions, todos, session summaries and a concise project brief, not raw speculation or every line of conversation.
A useful checkpoint states what was decided, who or what established it, when it applies, and what remains unresolved. A useful open question identifies the owner and next step. A useful project brief summarizes objectives, current status, constraints, key sources, recent decisions and immediate work without trying to replace the underlying files.
Central Brain's memory tools recall saved memory (memory_recall), append notes (memory_checkpoint), and replace the project brief while marking older notes as compacted rather than deleting them (memory_compact). Memory is stored as readable local files, so the consultant can inspect the record. Review and correct memory before relying on it; a durable mistake is still a mistake.
How does the start-of-session and end-of-session loop work?
Begin by recalling the project brief and relevant notes; end by checkpointing only the durable changes the next session needs.
At the start, confirm the client and project, review the brief, recall recent decisions and open items, and search the project files for the task at hand. During the session, distinguish a proposed idea from an approved decision. At the end, record established changes, update open questions and todos, and compact the brief when it no longer reflects the project.
This loop reduces repeated explanation, but it does not guarantee that a connected AI tool will call every tool or use every retrieved item correctly. Make the expected loop part of the project instructions and review material outputs against client sources.
What are the confidentiality and handoff limits?
Local project storage narrows the default boundary, but client confidentiality still depends on the device, folders, backups, accounts, transmissions, people and providers involved.
Keygen, which handles Central Brain licensing, receives licensing and device data, not documents. Connected AI tools receive the context they retrieve. Optional cloud connections and synced storage receive the data used with them. People or software with access to the workstation or storage may also access the project. Do not use sensitive or regulated client data until the engagement owner has approved the complete workflow.
For handoff, deliver only the materials the agreement requires, in an agreed format and through an approved channel. Explain which records are source documents, prepared output and memory notes. Central Brain does not by itself define the client's retention, deletion, confidentiality or records-management obligations. Review the canonical security guide before establishing the process.
What should a consultant validate before using project memory?
Validate separation, recall accuracy, source access, account boundaries, backup, deletion and the engagement's rules with non-sensitive test data first.
Create two test clients with deliberately different facts. Confirm that searches and memory recall stay within the selected projects and that project connections point to the correct folders. Try to provoke a mistake by switching projects between sessions. Verify that the visible project identity is sufficient for the working practice.
Then test backup and restore, removal of obsolete files, correction of an inaccurate memory, and closing an engagement. Document who approves deletion and which artifacts remain. For a file-search pilot within one approved client project, continue to the private business-file guide.
Validation checklist
What should you verify before relying on the workflow?
- Create one project folder, output location, index and memory set per client or engagement.
- Confirm engagement permission before adding any client file or sending any context.
- Record decisions separately from proposals, and keep owners on open questions and todos.
- Test two-client separation, correction, backup, restore and engagement closure.
- Document connected accounts, licensing, optional connections, access, retention and deletion.
Common questions
Can Central Brain keep each consulting client in a separate project?
Yes. A consultant can use a separate project folder, prepared output, local index, memory files, project brief and connection for each client or engagement.
What should a consultant save in project memory?
Save established decisions, verified facts, open questions, todos, session summaries and a concise current brief. Avoid storing unverified speculation as fact.
Does project separation automatically satisfy confidentiality obligations?
No. The consultant must also govern devices, folders, people, accounts, backups, transmissions, optional connections, retention and the engagement terms.
Can a connected AI tool mix up client context?
Any person or connected tool can be used with the wrong project or account. Separate projects reduce scope, but the consultant must verify the active project and review important outputs.
Next steps
- Security and data boundaries: what stays local and what a connection sends
- Central Brain product page: the canonical product and feature description