Zed is built for fast, agentic development. BrainOS adds the durable memory layer a coding agent needs when a project has history: decisions, blockers, plans, and next moves that survive the current session.
Zed memory should not just be old messages. For agentic coding, the useful memory is operational state. BrainOS stores that state locally and makes it available through MCP-compatible workflows.
Fast editor workflows make it easy to jump between tasks. They also make it easy for project state to drift. A coding agent can see files, but it may not know why a direction was chosen, what plan is active, or which blocker explains the current shape of the repo.
BrainOS gives Zed and other MCP clients a shared state layer, so the agent can inspect project continuity instead of reconstructing it from fragments.
BrainOS stores memory as typed project state, not a vague notes file:
Then add BrainOS to your MCP-compatible client config:
{ "brain-os": { "command": "npx", "args": ["brain-os"] } }
It helps when Zed or another coding agent needs to understand decisions, active plans, blockers, or what changed since the last session.
If the task is a single-file fix with no future consequence, a persistent memory layer is probably unnecessary.
The BrainOS pilot is open for developers testing agent memory in Zed, Claude Code, Cursor, GitHub Copilot, or any MCP workflow.