Design & Creativecreative brief memory across agency client work

Give Agencies Creative Brief Memory That Compounds Across Every Client Project

Agency creative briefs repeat structural patterns across clients. The same questions, the same dependencies, the same approval hurdles. AI tools used for brief generation see one project at a time. MemoryLake gives agencies creative brief memory that compounds across clients.

Get Started Free

Free forever · No credit card required

The problem: agency creative briefs don't compound across projects

The brief structure that worked for Client A's launch is rewritten from scratch for Client B's similar launch. The questions nobody thinks to ask in client kickoffs get re-discovered every time. The brief-writing skill of the senior strategist doesn't transfer to AI tools.

How MemoryLake captures agency brief memory

Brief template skill memory

Per-project-type templates with required elements.

Per-client context memory

Each client's approval style and constraints.

Reflection memory across briefs

Patterns of what works and what doesn't.

Cross-tool retrieval

Brief docs, project management, creative tools unified.

Get Started Free

Free forever · No credit card required

How it works for agency brief memory

Connect — Authorize agency tools and import past briefs.
Structure — Templates, client contexts, and reflections become typed memory.
Reuse — Every new brief retrieves relevant patterns and client context.

Before vs. after: agency creative brief memory

	Without MemoryLake	With MemoryLake
Brief structure quality	Variable	Memory-driven
Client-specific brief adaptation	Manual	Memory-applied
New strategist learning curve	Months	Memory access
Cross-project pattern detection	Manual review	Reflection memory

Who this is for

Marketing, creative, and design agency strategists at 5-50 person agencies — where brief quality directly affects creative output and onboarding cost is real.

Runnable demoabout 5 minutes

Try this use case on your own account

A design studio keeps its past client briefs in shared memory so the brief assistant can answer "when does Fernlight launch?" before the next brief is written. The briefs are Word files, and the review happened in comments: suggestions the team turned down were closed with Resolve and no reply. In Word they grey out; in the file only one flag changes. The index reads every comment as an ordinary line, a resolved one exactly like an open one, MemoryRouter attaches it, and the assistant answers with the idea that was turned down. The same files with resolved threads removed before upload come back right, and the open comments still get used.

What you will see

Three Word briefs (two with comment threads, 7 resolved and 2 open; one with none as the control) stored twice: as saved, and after drop_resolved() (about 40 lines of standard library that removes resolved threads and keeps open ones), one project and one boundary each; the assistant is router chat --boundary (gpt-4o-mini, temperature 0)
search --types document shows each comment as Commented [PN1]: …: all 7 resolved ones are in the as-saved index text, and the word "resolved" appears 0 times
Eight questions, 3 answers each, over 8 runs: on lines with a turned-down suggestion, as saved 2 of 144 right ("launches on Tuesday, November 3", "Yes, the budget has been raised to 60,000"), resolved dropped 144 of 144; the question answered by an open comment 24 of 24 under both
A checker call (Claude copies out what MemoryRouter attached) shows the resolved comment arriving with the brief 45 times in 48. Plus brief and threads subcommands and a local web UI

You need: a MemoryLake API key · the memorylake CLI · Python 3.9+

Run it

git clone https://github.com/memorylake-ai/memorylake-usecases.git
cd memorylake-usecases/resolved-comment-memory
export MEMORYLAKE_API_KEY=sk-…
python3 demo.py          # or: python3 web/server.py

Open on GitHub Read the walkthrough