What happened
Anthropic released Claude Opus 5.1 on July 28 with persistent agent memory across sessions. The model carries project context, preferences, and prior decisions without requiring users to re-upload context each time.
It is available to Max, Team, and Enterprise subscribers and via the API at $15 and $75 per million input and output tokens. Anthropic said beta tests with Cursor and Replit showed a 60 percent reduction in repeated context-setting. Memory scopes are configurable per project and per team.
Why it matters
Context re-entry is one of the biggest friction points in multi-day agent workflows. Persistent memory addresses that directly and could make long-running coding and research agents practical.
Configurable scopes also matter for enterprises that need to prevent memory leakage between teams or clients.
MintedBrain take
Persistent memory is powerful and risky. Define memory scopes tightly for client-facing work and audit what the agent retains before trusting it with sensitive project history.
Discussion
Sign in to comment. Your account must be at least 1 day old.