Glen

Glen is an organizational memory layer for teams running AI coding agents at volume. Add it as a plugin to Claude Code, Codex, or Cursor, and it captures agent sessions, Slack, GitHub, tickets, calls, and docs continuously, synthesizing everything into one queryable company store. Rather than you pasting context, Glen finds the prior work that bears on a task and steers your model with it automatically. The vendor reports equivalent work comes back faster and on fewer tokens. The setup pitch is short: one plugin install plus a one-time integration sync, and the vendor says full functionality is live in under five minutes. Only one person needs to connect the sources. Any agent, including ones you build yourself, reaches the store through an MCP server or command line tool; the web app, PR reviewer, and Slack bot come with the account. Beyond retrieval, Glen layers on collaboration features: a shared transcript library where teammates can hand off half-finished sessions mid-task, skills distilled from real team work that surface automatically when someone hits a covered task, and provenance tracking that keeps a queryable timeline of who decided what and when. A PR reviewer query clones of the agent that authored a line to answer why a thirty-second retry backoff exists, pulling in the incident, the Slack argument, and the originating ticket. The August 2026 benchmark tells the clearest story: replaying 100 merged PRs through Codex 5.5 with and without team memory produced 29% fewer tokens and 21% faster completion, quality at parity or better. Glen reads public material only (public Slack channels, shared transcripts). Compare it to simple RAG-over-docs tools: Glen is narrower on source breadth but deeper on agent-native context injection and cross-session handoff.