How do you found agentmemory so far? happy to help!

308 views

Add a comment

Replies

Best

Really like the direction so far , especially the focus on making memory actually useful for agents instead of just storing context endlessly . Feels practical , which is rare in this space.

One thing I noticed while trying different agent frameworks is that , memory , often becomes just more noise over time. Agentmemory feels more intentional compared to a lot of tools in this space. Curious how you're thinking about memory decay or relevance ranking as conversations get longer?

there is decay algo already.
nice to hear this.

Tried it briefly , feels clean and easy to get stared with. How you're handling memory relevance over time , especially when the context starts scaling up ?

it's just KV store.

Got it , KV store is a clean starting point and keeps things straightforward.

Took a quick look , seems straightforward and usable. Curious how you're keeping memory relevant without it becoming noisy over time.

it has more practical implications than simply storing memory, which sometimes create a un-intended noise in tools.

Been using it for 2 weeks, and I definitely see improvements.

One thing I wish AgentMemory had -- team collab in some way. AgentMemory should have some way to export relevant memories to filesystem which can be committed to git. Also, some way to import them. This is something we want to try out internally in a few weeks once our devs all have used it some time and memories have evolved.

I think skillkit.sh can do this in agentmemory
Haven't tried yet... currently have supermemory and obsidian hybrid system for Hermes. I'm tempted to use Agent memory without complicating my current system. any suggestions?
agentmemory is simple to use and you can simply ask agent to transfer obsidian data into agentmemory state. it should work.

I started using it to help with the memories of how my product has pivoted over time. I'm pre-revenue and it's CONSTANT pivots and cursor by default was impossible to keep tidy. Slight variations in direction looked identical in the landing pages, app, backend, and I needed something that tracked the conversations more deeply.

I backfilled agent memory on my past month's cursor Agent transcripts and asked it some questions to see how well it did, and it was surprisingly accurate. Picked up on things that I move away from, decided we're badly done, etc, which aren't immediately obvious if you look at my codebase despite my diligence with updating documentation and labeling commits.

So far, so good.

Wow 👌 👏

HELLO Really promising so far — I like that it tackles one of the biggest pain points with coding agents: losing useful project context across sessions without bloating the context window. The searchable memory + visibility into what the agent remembers feels especially valuable. 👏

Thanks Rohit for the tip off 😎

Agent memory is the layer everyone needs but nobody wants to build themselves. The teams getting the most out of AI pipelines today are the ones who've solved persistent context. Curious how you're thinking about structured vs. unstructured memory for content-heavy workflows.

glad to hear this
12
Next
Last