Google ships Gemini Code Assist Enterprise with persistent codebase memory
TL;DR
Google announced Gemini Code Assist Enterprise with a persistent codebase memory layer that survives across sessions and pull requests. The agent reads commit history, tracks open PRs, and keeps team-specific style rules between calls. Pricing starts at 19 dollars per developer per month.
What changed
Google announced Gemini Code Assist Enterprise with a persistent codebase memory layer that survives across sessions and pull requests. The agent reads commit history, tracks open PRs, and keeps team-specific style rules between calls. Pricing starts at $19 per developer per month with volume discounts above 100 seats.
Why it matters
Persistent memory across sessions is the gap that has separated coding agents from "smart autocomplete" for two years. Without it, every AI coding tool starts each conversation cold and burns prompt budget re-explaining the codebase. Gemini Code Assist Enterprise solves the cold-start problem at the team level, not the individual level.
What to watch for
Whether Cursor and Windsurf ship comparable persistent-memory layers in response, or pivot to a different differentiator. Real-world latency on the memory-retrieval step at codebase scale (10k+ files). Privacy-mode pricing for teams that cannot send code to Google for indexing.
Who this matters for
- For Vibe Builders: Gemini Code Assist Enterprise remembers your codebase between sessions, so the cold-start cost of explaining your project to the agent every morning goes to zero.
- For Developers: persistent codebase memory at $19 per dev per month closes the main gap between Gemini and Cursor + Windsurf for team deployments.
What to watch next
Persistent memory is the feature that turns coding agents from convenience into infrastructure. Once your team has 6 months of accumulated codebase context inside Gemini, switching tools costs you that institutional knowledge, not just the workflow change.
The $19 per developer per month price is aggressive: Cursor Business is $40, GitHub Copilot Enterprise is $39. Google is using their model + infrastructure cost advantage to win on price at the enterprise tier specifically. Worth tracking the seat-counts at large customers in Q3 earnings.
by Harsh Desai
About Gemini
View the full Gemini page →All Gemini updatesMore from gemini
- Model ReleaseGoogle releases Gemma 4 under Apache 2.0 for agentic workflows
Developers access Gemma 4 on Google Cloud with Agent Development Kit to build reasoning and agentic apps alongside proprietary Gemini.
- FeatureGoogle Adds Gemini Auto-Browse to Enterprise Chrome
Enterprise Chrome users get Gemini-powered auto-browse for tasks like research and data entry. Automates workplace workflows.