Solving the Agentic Memory "Hoarding" Problem with Biological Pruning

[sachitrafa]

Agentic AI memory is a hot topic, but we are looking at it the wrong way.

I’ve always viewed agentic memory as a Dynamic Programming problem. We should be storing sub-problem results to reach a final solution without recomputation. However, current "hoarder" architectures treat memory like an infinite hard drive, saving every raw log. This leads to massive token bloat and a collapsing signal to noise ratio.

In most cases, we don’t need the results of every sub problem forever. We need the relationships between facts and a way to prune the noise.

That is why I built YourMemory.

It is a biologically-inspired persistent memory engine (available as an MCP server and API) that mimics how human storage works. Instead of just being another vector database, it uses an Auto Pruning mechanism based on the Ebbinghaus forgetting curve.

Why this approach?

1. Relationship Mapping: It stores the links between entities, so when an agent recalls one fact, the relevant context follows naturally.
2. Resource Efficiency: By letting stale, low signal data decay, we keep the context window sharp and save significant tokens.
3. Local First: Built to run on your infrastructure (Postgress/DuckDB). Your data stays yours.

The Benchmarks

I tested this against the LoCoMo dataset (1,534 multi session QA pairs) to see how it stacks up against traditional memory layers. The results show that "smarter forgetting" actually leads to better recall:

By focusing on high signal data and pruning the rest, we achieved nearly 2x the accuracy of established tools.

I’d love to get your thoughts on this "forgetting" philosophy. Are we over-relying on raw context? How are you all handling memory bloat in your agentic workflows?

Check out the repo here: https://github.com/sachitrafa/YourMemory
Also a Star on my repo will be appreciated !

#AI #AgenticAI #MachineLearning #OpenSource #MCP

[armorer-labs]

Memory pruning gets talked about mostly as a context-efficiency problem, but for agents it also feels like a trust-management problem.\n\nIf a system hoards too much memory, two bad things happen at once:\n- the context gets noisy and expensive\n- stale or weakly trusted information keeps showing up with accidental authority\n\nThat second part matters a lot. A remembered tool result, scraped page, or low-confidence observation can keep influencing decisions long after it should have decayed. So pruning policies probably need to consider not only relevance, but also provenance and authority: where did this memory come from, how old is it, and what kinds of downstream actions should it still be allowed to influence?\n\nIn other words, not every useful memory should have the same half-life.