Aurra
Memory for AI agents that ages gracefully
Start new thread
Akshay Trivedi

Aurra - Memory for AI agents that ages gracefully

by•
The memory layer for AI agents. Persistent, source-cited, and honest about what it doesn't know. Install in 5 lines of Python.Most agent memory layers lose track of what changed. Aurra adds bi-temporal versioning to every memory: when it was true, when it stopped, what replaced it. Now with optional LLM auto-detection of supersession. Free tier, both Python and JS SDKs.

Add a comment

Replies

Best
Akshay Trivedi
Maker
📌
Founder here, with the team. Three things worth saying: 1. This is targeted at developers building agents on top of Anthropic, OpenAI, or open models. If you've ever shipped an agent that "forgot" something the user just told it, or remembered an outdated fact as if it were current, that's the failure mode Aurra solves. 2. Free tier is 500 writes/mo with no credit card. Realistic for prototyping. Pro is $99/mo for 500k writes which covers most production agents. 3. The Level 2 auto-supersession that just shipped is in beta. The API works, the SDKs are out, but live classification is gated behind a server flag for ~3 days while we validate against production traffic. We're not pretending it's GA. The blog post explains exactly what's flagged and why: https://aurra.us/blog/level-2-au... Bonus: we open-sourced the eval methodology + 121-case test set: https://github.com/aurra-memory/... Happy to answer anything.