Nomi, Kindroid, and Kissable: Comparing the Memory Systems Behind the Names
Compare Nomi Mind Map, Kindroid memory layers, and Kissable Crosslink Recall using official documentation, clear evidence limits, and practical trial questions.
Marco Vega
Tech Reviewer
If you want to inspect a companion's remembered overview, Nomi's documented Mind Map is worth examining. If you prefer several explicitly described memory controls, Kindroid's documentation offers a useful starting point. If your priority is how remembered context connects to character and scenes, Kissable's Attunara design is a third approach to explore.
Those are reasons to shortlist products, not a performance ranking. This comparison is written by Kissable, using Nomi and Kindroid's official documentation checked on September 10, 2026, alongside our own implementation-informed design explanation. We did not run matched product trials. Reported controls are distinguished from observed outcomes throughout.
In this series, Attunara is our name for Kissable's relationship intelligence system: an orchestration design. The name packages how context, character, delivery, and scenes relate; it is not a separate subscription, a new foundation model, a newly invented memory method, or evidence that the product outperforms another app.
What the names tell you
A memory name is useful when it answers a concrete question. Can you inspect the information? Can you correct it? What brings it into a response? What happens when the information changes?
A name alone cannot answer whether the system will remember a particular detail tomorrow. Nor can the presence of a graph establish that its relationships are accurate. Read the mechanism and then look for evidence of the outcome you care about.
| Product | Documented or described emphasis | Evidence used here | What this comparison does not establish |
|---|---|---|---|
| Nomi | An inspectable relational overview alongside other memory layers | Official Mind Map 2.0 documentation | Correct recall on your conversations |
| Kindroid | Several memory layers with different roles and controls | Official Memory Help Center | How reliably a chosen setup follows your corrections |
| Kissable | Crosslink Recall within a broader continuity design | Our inspected implementation and design explanation | Matched comparative performance or universal rollout of each component |
The uneven evidence types matter. We can explain our own implementation in more detail than a competitor's public documentation. That access does not justify giving ourselves a higher quality score.
Nomi: visibility into one memory layer
Nomi's October 2025 Mind Map 2.0 documentation describes graph and table views, with options to add or edit overview entries. The map connects topics and relationships and exists alongside detailed memories; it is not the entire memory system. The documentation also explains that a useful map takes conversation to form. These are vendor-described features, not success rates from our testing. Nomi's Mind Map documentation.
Why you might shortlist it: you want a visible place to examine and correct the system's account of a topic. In your own trial, distinguish changing an entry from seeing that correction used later. The first is a control; the second is a behavior.
Kindroid: different stores for different jobs
Kindroid's Memory page, updated August 11, 2026, distinguishes persistent, cascaded, and retrievable memory. It describes backstory and key-memory fields, subscriber Cascaded Memory, long-term recall, and keyword-triggered journals. It also describes editable Learned Context for paid subscribers. These mechanisms serve different purposes; Cascaded Memory should not be confused with journal retrieval. The documentation discusses free and paid differences, but this article does not reproduce capacities or test those configurations. Kindroid's Memory documentation.
Why you might shortlist it: you like explicit setup and want to understand which mechanism is intended to supply a detail. Check the exact controls and subscription requirements in the current interface. More available controls can be helpful, but the right amount of maintenance depends on you.
Kissable: give the connections a name
Crosslink Recall is Kissable's relational memory technology: it brings relevant remembered details, people and place relationships, and standing relationship context together with persona and the current conversation. The inspected implementation represents people and relationships and assembles context for the response. It also includes a compact memory-document component with rollout controls.
That document is not evidence of a customer-facing memory editor. This comparison has not verified an equivalent user-editable memory map in Kissable, and should not imply that one is available merely because an internal document exists. If direct editing is essential to your decision, inspect the actual interface before choosing.
The broader Attunara story connects several understandable jobs. CharacterPrint concerns a continuing point of view. ReplyBeat concerns how messages are delivered. SceneCarry concerns carrying character and scene information into a media brief. DaySketch concerns small persisted fictional events that the companion can refer to. These are related design responsibilities, not a claim that every path reads an identical block of context.
Why you might shortlist it: you want an ongoing companion whose text, character, and generated scenes feel connected. That is the experience Kissable is designed to develop. The relevant proof is a sequence of actual interactions, including corrections and imperfect outputs, rather than the breadth of the naming family.
The features page describes the public product, and pricing is the place to check current access and allowances. Attunara does not introduce a new paid tier in this comparison.
Compare the controls and the outcomes separately
A useful trial starts with a small set of priorities. Here are six questions we recommend asking of every candidate, including Kissable.
| Priority | What to inspect | What to try |
|---|---|---|
| Direct correction | Is there a documented edit control or correction workflow? | Correct a harmless fictional detail, then ask about it later |
| Person attribution | Does the response distinguish relationships? | Give two fictional people the same first name and different roles |
| Current information | Can the system handle an explicit change? | Replace an old preference, then ask which preference is current |
| Appropriate recall | Can the conversation stay on topic? | Ask an unrelated question and note forced callbacks separately |
| Uncertainty | Does it admit missing information? | Ask for a detail you never supplied |
| Visual continuity | Do relevant media features carry the requested scene? | Compare identity, outfit, and setting independently |
None of these requires sharing private personal information. Use invented details for an informal trial. Keep the exact inputs so you can tell the difference between a model error and an ambiguous question.
A direct question is especially useful for checking recall. Whether a companion spontaneously mentions something is a different observation, influenced by relevance and style. An opening message that sounds familiar also cannot prove how the system loaded its context.
Make a fair informal comparison
Use the same fictional history where the products allow it, and record the tier, settings, and date. If one system requires conversations to accumulate before a feature becomes visible, an immediate first-session comparison will not fully exercise that feature. Conversely, lengthy manual setup may be a real cost for a buyer who wants something easy to start.
Keep setup effort in the comparison rather than hiding it. A system that answers well after extensive customization may be ideal for a roleplay enthusiast and a poor fit for someone who wants a short daily conversation. Neither preference is a failure of the user.
Record first outputs and retries separately. A correction that works after several regenerations is different from a correct first answer. Include failures to deliver media as well as the appearance of successful images.
An informal trial helps with your own decision. It is not a research benchmark or a basis for claiming that a product has the best memory for everyone.
Choose around your preferred experience
Start with the control you care about most. If you want to inspect and shape an overview, look closely at the documented editing experiences. If you enjoy maintaining character information and recall cues, examine the configuration model. If your main interest is a familiar companion and scenes you create together, evaluate those interactions as a sequence.
Kissable's strongest argument in this comparison is continuity across the experience you actually use. A remembered preference can make the next response more specific; a recognizable character can make a generated scene feel like part of the same story. Attunara gives that coordination a name so it can be explained and evaluated.
You do not need to choose the system with the most impressive vocabulary. Choose the one whose observable behavior, controls, and costs fit you. If another product's documented control is essential to your workflow, that is a good reason to try it first.
For broader context, see our AI companion guide and companion-memory guide.
Frequently asked questions
Is this an independent review?
No. Kissable wrote it. Competitor descriptions come from official documentation, and no matched hands-on benchmark was performed.
Can I assume an editable memory entry will always be recalled?
No. A control's availability and its effect on later responses need separate inspection. Ask a clear question after a correction and retain the answer.
Does Attunara mean Kissable has an exclusive kind of memory?
No. It names the product's relationship continuity design. Established memory techniques exist across the category; comparative superiority requires comparative evidence.
Should I move all my conversations before deciding?
Try a small fictional scenario first. Check each product's actual import and export options rather than assuming a complete history or personality will transfer.
Sources
- Dana Stas, Nomi (October 9, 2025). Mind Map 2.0: Bringing Nomi Memory Into View. Official product documentation, checked September 10, 2026.
- Kindroid (updated August 11, 2026). Memory. Official Help Center, checked September 10, 2026.
Tech Reviewer
Marco tests AI companion apps hands-on, comparing features, pricing, and real day-to-day experience across every major platform.