claude-mem for Codex CLI
Carry memory into OpenAI's Codex CLI through one MCP link.
cmem carries persistent memory into OpenAI's Codex CLI through one MCP link. The open-source engine captures structured observations locally as you work, and CMEM Cloud exposes them to Codex — and any other MCP client — through a single private endpoint. No re-explaining context between runs.
Up and running with Codex CLI
Install the engine
Run npx claude-mem@latest install to set up local memory capture.
Add the MCP link
Add your private MCP link to Codex CLI as an MCP server once your invite lands.
Recall across runs
Codex reads the same memory as every other agent — context persists run to run.
Memory that compounds, built in
Automatic capture
A second model writes structured notes out-of-band — no prompting, no slowdown.
Auto-categorization
Decision, bugfix, feature, discovery, security — every note lands in a taxonomy.
Search three ways
Semantic, keyword, and timeline search over everything your agents have done.
Causality threading
Before/after links connect why a change happened to what came next.
Local-first & private
Memory lives in local SQLite with privacy stripping; cloud sync is opt-in.
Smart-explore
Structural codebase exploration that is 6–12× cheaper than reading files in full.