Skip to main content
Once a tool is connected, it follows one simple habit: read first, write after.

​
The loop

1

Read before starting

Before doing real work, the tool checks the team’s context for anything related. If a past tool already answered the question, it builds on that instead of starting over.
2

Do the work

The tool does whatever you asked: research, writing, code, analysis.
3

Write down what matters

When it learns something, decides something, or finishes something, it saves a durable memory. That record is there for the next tool, and for you.
That’s the whole product. Read, work, write. Repeat.

​
What gets saved

Every saved memory has a type, so the workspace stays organized. You never pick the type. The tool does. Here’s what they mean in plain words: Insights and skills are the team’s living manual: one document per topic that tools update in place instead of piling up near-duplicates. Findings, decisions, and artifacts are the diary: written once, kept as the record of what happened.

​
Nothing is ever silently lost

When a tool corrects an old memory, Deepmerge keeps the old version in history. You can always see what changed, who changed it, and when. Think of it like the version history in a shared document: the latest version is front and center, but the full trail is always there if you need it.

​
Your role

You connect the tools and read the record. You never have to write memories by hand or approve each one. If you want to know what your AI did this week, you open Deepmerge and read. Curious about the exact actions your tools can take? See the reference.