> For the complete documentation index, see [llms.txt](https://vayl.gitbook.io/vayl-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://vayl.gitbook.io/vayl-docs/documentation/why-vayl/why-your-agents-memory-returns-stale-facts.md). # Why your agent's memory returns stale facts **Agent memory returns stale facts because most memory systems *****append***** instead of *****reconcile*****.** When a value changes, the old one stays searchable, and retrieval can surface it as current. The fix isn't a bigger model or more memory — it's memory that retires the old value on write and keeps exactly one active value per fact. ## The root cause, precisely Additive and vector memory (and today's built-in agent memory) treat every statement as a new item. Update a customer's plan four times and you have four `plan = …` memories, all timestamped, all retrievable. At read time the model sees several equally-plausible "current" values and guesses. Under sustained churn the silently-wrong rate *rises* — the more a fact changes, the worse it gets. ## Why "just prompt it better" doesn't hold Reconciliation-by-prompt depends on the model noticing the contradiction every time, across a long context, forever. Vayl makes it *structural*: the engine enforces one active value per slot, so correctness stops tracking model strength. A cheaper model can't produce two live contradictory values because the store won't hold them. ## The three shapes of agent memory * **Additive / vector** — simple, broad recall; **goes stale** under change, and can't cleanly model removal. * **Temporal graph** — reconciles and answers multi-hop, but heavy (a graph database plus many LLM calls per fact), and removal is often an afterthought. * **Reconciling (Vayl)** — supersede / retract / flag on write; one active value per slot; history kept; \~2 LLM calls, one SQLite file, no server. ## How to fix it today Point your MCP client at Vayl. `remember` extracts and reconciles; `recall` answers from the active set or says **"I don't know"** rather than guessing; `history` shows what changed. No migration off your existing RAG — Vayl handles the *changing* facts; your corpus tools handle the static ones. ## Next steps

:bolt: Quickstart

Point an MCP client at Vayl and store your first fact./pages/416edc1381c20849bc41111b02e482a722a49b94

:diagram-project: How Vayl works

The reconciliation model end to end./pages/OKgqCYi8fPKbjoks0QMg

:database: Memory tools

remember, recall, history, and the rest./pages/atmh20zKFo3dWE6nALWX
--- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://vayl.gitbook.io/vayl-docs/documentation/why-vayl/why-your-agents-memory-returns-stale-facts.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.