Supermemory alternatives
Checked against Supermemory’s own documentation on
Supermemory is context infrastructure rather than memory alone. It ingests from connectors such as Slack, Notion and Drive, with document retrieval bundled in.
That breadth is the reason to pick it and the reason people sometimes want something narrower. WOS is the narrow one, and it is ours: memory and nothing else, $0.0001 a request and no seats.
- +10.3LongMemEval-S95.7% against Supermemory’s 85.4%
- 95.2%recall@5 across 70 language pairsSupermemory publishes no figure for this
- 2,484tokens per query at 2.2M memoriesmeasured with a tokenizer, not estimated
Where WOS comes out ahead. Where it does not is in the table below.
Supermemory and WOS
| Supermemory | WOS | |
|---|---|---|
| What each one has published | ||
| LongMemEval-S | 85.4% their published figure | 95.7% Tablet 2 read by Claude Opus 5; 93.7% read by GPT-5.6-sol |
| LoCoMo P@1 | 59.7% their published figure | Not run |
| Latency | Sub-300ms at the scale they state | ~320ms median engine round trip; 393ms at 2.21 million memories |
| Also published | Temporal 82.0%, updates 89.7% their LongMemEval-S breakdown, plus MemoryBench, the open benchmark they maintain | BEAM-1M 67.5%, languages 95.2% Tablet 2; recall@5 over 70 language pairs |
| Source | supermemory.ai/docs/memorybench | how we benchmark |
| How it is built | ||
| Stored as | Content from connectors and chat | Kept, not rewritten |
| A model reads it | Yes, at extraction | Never |
| Organised around | Connected sources and their documents | One store per end-user |
| Retrieval | Retrieval with document RAG bundled | Not published |
| Correcting a fact | Worth checking in their docs | supersede, with lineage |
| Licence | Not open source | Closed source |
| Self-hosting | On the higher tiers | Case by case |
| Metered by | Storage, extraction, search, operations | Tokens, plus a flat fee per request |
- Both columns are self-reported: a placement, not a verdict. Some rows go to Supermemory.
- Across languages is a grid of 70 store-language by query-language pairs.
What WOS does
Against Supermemory the difference is width. A platform that covers connectors, documents, retrieval and operations is fewer vendors and more surface. A memory-first system spends all of its surface on the behaviours that only matter for memory.
For your AI Won is the part of the product addressed to the model rather than to you, and it carries no charge. Your agent can ask how much of its own memory has been rewritten before it leans on a fact, and the answer is the same size for a hundred memories or a hundred million. the page written for an arriving agent.
- Who said itA group conversation does not collapse into one undifferentiated voice. Register a speaker once and memories carry attribution afterwards.
- Facts that changesupersede replaces a memory so the outdated version stops surfacing, and lineage reads the whole chain of revisions back. Nothing is silently deleted.
- A photograph as a memoryAn image can be the memory itself and is found from a sentence describing it, in any language, rather than from a filename.
- Repeated queries bill a tenthCaching is a field on the call and a write invalidates it instantly. One meter, and the cheap path is cheap.
- Writes are safe to retryEvery write takes an Idempotency-Key, so a retry replays the first answer instead of storing the same turn twice.
- Calls the model makes for itselfWon is the part of the product addressed to your agent rather than to you, and it carries no charge. revisions is the first of them: the model can ask how much of a memory has been rewritten before it leans on it, and the answer is the same size at a hundred memories or a hundred million.
- Multi-hop without a modelEngrams search and then expand around what they found: deep recall, timelines, drafts. No language model runs in that loop, so a question that takes two steps is one call and there is no per-hop bill.
Where Supermemory is the better answer
- Connectors to Slack, Notion and Drive that work without you writing an ingestion layer
- Document retrieval bundled in, so one dependency covers files and memory together
- A single vendor for context of every kind, which is genuinely fewer moving parts
- What WOS costs you. It is closed source and hosted, so running it on your own hardware is a conversation rather than a download. There is no knowledge graph, so a question that is really about validity intervals is better served elsewhere. And it is the newest name on this page, with a smaller ecosystem than the ones above it.
- What it buys you. Nothing runs a language model over what you stored, so the privacy review is one sentence rather than a discussion. A query comes back in about 320ms, and the context it hands over was measured at 2,484 tokens over 2.2 million memories. Repeated queries bill a tenth. And it behaves the same in every language: recall@5 of 95.2% over 70 store-language by query-language pairs.
Why teams look at something else
| You may only need one of the things it does | If your problem is conversational memory alone, a platform that also does connectors and documents is more surface than the problem calls for. |
| The memory-specific behaviours are worth checking first | Speaker attribution, and correcting a stored fact without losing its history, are worth confirming in their docs before committing. |
| Four meters instead of one | Storage, extraction, search and operations are billed separately, so forecasting a month means modelling four things. |
| Self-hosting sits on the higher tiers | If running it yourself is a requirement rather than a preference, that requirement carries you to a particular tier. |
- None of these are defects. Each follows from a decision that is right for somebody.
Bringing memory over
Text is text, so the move itself is unremarkable: create one store per end-user and push their history through bulk-store. The question worth answering before you start is which half of the platform you were actually using. If connectors and document retrieval were doing real work, replacing the whole thing with a memory API leaves a gap you will have to fill yourself, and keeping both is usually cheaper than rebuilding one.
The calls are in the documentation: create a store, bulk-store, and the rest of the endpoints. A key takes a few minutes in the console.
Questions
- What is a memory-first alternative to Supermemory?
WOS. The whole surface is memory: speakers so a group conversation does not collapse into one voice, supersede and lineage so a corrected fact keeps its history, images that can be found from a sentence, and multi-hop retrieval that runs without a language model.
- How is WOS metered by comparison?
WOS has one meter: input and output tokens per million, plus a flat fee of $0.0001 per request, at the same rate at any volume. There is no separate charge for storage, extraction or operations, and no seats.
- Is Supermemory open source?
It is a hosted platform, with self-hosted deployments available on the higher tiers rather than as an open download. WOS is hosted only, so if you need the source neither of these is the answer.
- Do I still need document retrieval?
If your product answers questions about files, yes, and WOS will not replace it. The two are complementary: per-user memory for what this person told you, document retrieval for what your organisation knows. Supermemory covers both in one place, which is the argument for it.
The other options
- Mem0 Extract facts with a model as you write. Right for you if you want conversations turned into clean facts, and the widest ecosystem.
- Zep Build a temporal knowledge graph. Right for you if your questions are about when something was true, not just whether it is true.
- Letta Let the agent manage its own memory. Right for you if you want the agent to own and edit its own memory, inside their runtime.
- Cognee Run a pipeline over documents into a graph and vectors. Right for you if what you are remembering is a document corpus, and you want to own the pipeline.
- Vector DB Build it yourself on a vector database. Right for you if you are prototyping, or you want no vendor in the path and will build the rest.
- Checked against Supermemory’s own documentation on September 1, 2026.
- We quote nobody else’s prices, and use only figures each project published itself.
- Out of date or unfair? Tell us and it gets corrected.
You need the same store to work in many languages, and a per-query cost that does not drift as history grows. If that is not your case, one of the projects above will serve you better.
Three calls: create a store, write a memory, recall against it. No subscription, balance starts at $5. Get an API key or read the docs.