Claude Code stops forgetting between sessions: claude-mem captures tool observations and writes semantic summaries for later sessions
Synopsis
claude-mem gives Claude Code persistent memory across sessions: it automatically captures tool-usage observations during a session, generates semantic summaries, and makes them available to future sessions, so Claude keeps continuity of knowledge about a project after sessions end or reconnect. Installation runs through npx claude-mem install or the plugin marketplace, and the installer configures everything before asking the user to sign in and choose a memory provider.
Interpretation
Context in Claude Code can now carry across sessions instead of restarting when a session ends or reconnects. Previously, context was not retained between sessions, so a new session had no direct access to the prior round of tool usage. The capability ships as a plugin; after installing and restarting Claude Code, context from previous sessions automatically appears in new sessions.
Memory is produced by first recording tool-usage observations and then compressing them into semantic summaries. The mechanism is described as a persistent memory compression system built for Claude Code. The install entry point is npx claude-mem install, and the plugin can also be added and installed from the marketplace inside Claude Code, taking effect after a restart.
The install flow separates configuration from account sign-in, and the user picks a memory provider after signing in. Sign-in uses an email magic link with no card required and includes a 30-day free trial. When the trial ends, memory automatically falls back to the Anthropic plan unless the user subscribes; sign-in can be skipped with an explicit --provider flag, by disabling the online option, or in CI and non-interactive shells.
Perspective
The demonstrated scope is the installation and memory-recall flow in Claude Code and the host environments the material lists, for developers using those coding tools. The material states no next experiment or evaluation plan.
The material states that memory automatically falls back to the Anthropic plan when the trial ends unless the user subscribes; the Grok Bot awareness pilot only appends dated entries to that bot's monthly log, does not write profile.md, user memory or project memory, and can be disabled with an environment variable.