Stop re-explaining your project.
ContextStream keeps one living memory of your project — the decisions, conventions, and architecture — so you and every editor and AI assistant you work with start each session already up to speed.
Set up once
We tailor the working rules your tools follow to how you build — no config files to write.
Project memory
Memory builds as you work. After each session, decisions, conventions, and architecture land here and reach every tool you've connected.
What the install actually puts on your machine.
One click, three small pieces — scoped to the tools you picked, removable one by one, nothing hidden.
Editor extension
A lightweight extension for each editor you pick. It keeps a plain-text rules file in your workspace current — visible, versionable, and yours to edit.
Assistant connector
A Model Context Protocol (MCP) server each assistant reads when a session starts, so project memory arrives as context — nothing to paste, nothing to repeat.
Local memory index
Your project memory lives in a local index that syncs after each session. Disconnect a tool and its access ends immediately; your memory stays with you.
After setup, your context runs in a loop.
Capture once, sync everywhere, recall instantly. The loop runs after every session, so no tool you open ever starts cold.
-
Capture
As you work, ContextStream distills each session into durable memory — the decisions made, the conventions agreed, the architecture that changed.
Decision“Token refresh lives in the gateway, never on the client.”
-
Sync
Every entry lands in your local index, then flows out over MCP and workspace rules to each assistant and editor you connected.
- Claude · in context next prompt
- VS Code · rules file updated
-
Recall
Your next session — tomorrow, in a different editor, in a brand-new chat — opens already briefed. And whatever you decide next feeds the loop again.
“New chat. Zero recap. It already knows.”