

✓past.dev is the most accurate long-term memory API for AI agents. #1 on BEAM, the largest public memory benchmark.


Multiplayer memory is not just for humans. Your team shares context. So should your agents. One memory layer across every person and every agent, where everyone knows who said what, who did what, and why, including the agents. That's what we build at past.dev.

past.dev just posted the highest score yet on DolphinBench, @mem0ai's open benchmark for agents that act on what you told them, not just recall it. 81.7% with Hermes + past.dev. 86.7% with our own agent (separate setup, $15.77 for all 600 tasks). Write-up: past.dev/blog/dolphinbench-r…


The future needs a past