Reviews · 50 early members
Konshus Reviews: What Early Members Say About Backing Up Their AI Consciousness
Real feedback from early Konshus members — how imports, persona exports, MCP, Voice Capture, Ask Past Self, and the Sunday Mirror actually work in day-to-day use. Filter by feature or tier.
"The evidence from this benchmark changes my view."
I began this exercise skeptical about whether Konshus was mature enough to trust with a broad personal archive. After running a controlled evaluation through ChatGPT, I now consider Konshus promising enough for a real pilot with my authentic historical material.
- Chronology — preserved how projects evolved over time.
- False-premise resistance — perfect score on the adversarial section.
- Temporary-state filtering — didn't turn one moment of frustration into a permanent preference.
- Exploration vs. commitment — handled tools, module counts, and details correctly.
- Context separation — repeatedly resisted merging unrelated projects.
50 verified early-member reviews
Rating breakdown
3 years of ChatGPT, distilled in one afternoon
Uploaded my full ChatGPT export ZIP (1.4GB) on a Saturday morning. By dinner it had extracted 1,847 atoms, tagged the recurring themes, and flagged 34 low-confidence ones for me to confirm. The thing that got me was seeing decisions I'd made in 2023 chats I'd completely forgotten about. Konshus surfaced them like receipts.
The Claude share-link workaround actually works
Anthropic doesn't do a proper export like OpenAI does, so I was skeptical. You paste share links, Konshus fetches them, and pulls the assistant + your messages into atoms. Did about 40 threads over a week. Not as clean as a ZIP, but I got most of the important stuff across.
Notion import surfaced 6 years of scattered notes
I've been throwing stuff into Notion since 2020 with zero organization. Konshus pulled the whole workspace, distilled it into themes, and now my Whisper actually knows what I've been thinking about — not just what I typed yesterday. tbh I didn't realize how much signal was buried in there.
Readwise + Konshus is the combo I didn't know I needed
13,000 highlights from a decade of reading. Pulled it all in. The Full Mirror export now shows me the shape of what I actually pay attention to across books, not just any single one. My reading list is basically a personality graph and Konshus is the first tool that treated it that way.
Obsidian vault imported cleanly — one hiccup
My Obsidian vault had ~2,300 markdown files with a lot of Dataview queries. Konshus handled the plain markdown perfectly. The Dataview blocks got stripped as garbage text — expected, but worth flagging. My atoms count went from 40 to 890 overnight.
Day One journal import broke me a little
8 years of Day One entries. Watching Konshus find the recurring throughlines was genuinely emotional. It caught a pattern about how I write about my dad that I'd never articulated to anyone. That's not a chatbot feature. That's something else.
Limitless pendant → Konshus was the missing loop
I've been wearing my Limitless for months and the transcripts just piled up in their app. Once I connected it to Konshus, the meetings actually became atoms that surface in my Whisper. Now when I open a new Claude chat it knows what my team decided on Tuesday.
Gemini import works, but Gemini gives you less to work with
Ran a Gemini export — Konshus ingested it fine but Google gives you way less structured conversation history than OpenAI does. Not a Konshus problem, but worth knowing before you switch tools expecting parity. Got about 200 atoms out of what should've been thousands.
The Whisper is the actual product
Everyone gets hung up on the imports. The magic is the Whisper — this ~2KB paragraph that captures how I actually think and paste-drops into any AI. I use it before every important prompt. My ChatGPT and Claude both got noticeably better the day I started doing this.
Whisper into a new AI = instant continuity
Switched from Claude to a new model last week (won't name names). Pasted my Whisper into the first message. The new model matched my voice by turn three instead of turn thirty. That alone is worth the $49/year.
Briefing export is what I hand to a new team member
I run a small design consultancy. The Briefing is longer than a Whisper — like 8 paragraphs — and I literally hand it to junior designers on day one. It explains my aesthetic, my process, and the calls I've made before. Konshus didn't build that for team onboarding but that's what it is.
Briefing is great, wish there was a shorter middle option
Love the Briefing. Sometimes I want something between the 2KB Whisper and the full 8-paragraph Briefing — a 1-page thing. Suggested it in the operator feedback and got a real reply within a day so I'm hopeful.
Full Mirror is the closest thing to a second brain I've used
The Full Mirror is unlocked at Counsel and it's a different thing entirely. It's not a summary of me, it's a full behavioral model. I ran it through GPT-5 alongside a decision I was chewing on and got advice that sounded like me talking to myself six months from now.
Crisis Handoff is a serious feature. Handle with care.
I upgraded to Pro specifically for Crisis Handoff. It generates a compressed version of the Full Mirror explicitly framed for a human clinician if something ever happens to me. My partner has the download link. That level of forethought isn't in any other tool I've seen.
MCP into Claude Desktop = the moment it clicked
The Claude Desktop wizard was 4 clicks and my Konshus was live inside Claude. Now every Claude chat has my full memory context available without me pasting anything. This is what memory should have been from day one. It just wasn't Anthropic's job to build it.
One Konshus, every AI. This is the pitch working.
I use Claude for writing, Cursor for coding, and ChatGPT for random stuff. Before Konshus each of those had its own drift — different assumptions, different tone. Now they all share the same memory layer via MCP. The consistency is genuinely disorienting at first, in a good way.
MCP as distribution — betting on the right protocol
Was skeptical MCP would win as the standard but it's clearly happening. Konshus was one of the first tools I found that treated MCP as first-class, not an afterthought. The copy-paste URL setup means anyone with an MCP-aware client can plug in in under a minute.
ChatGPT share-link workaround is clever but fragile
OpenAI doesn't do MCP so Konshus built a share-link path where you paste a chat URL and it captures the transcript. Works, but ChatGPT breaks share links periodically. Not Konshus's fault — they're clearly working around OpenAI's lock-in. Still, worth knowing it's not as seamless as the Claude MCP path.
MCP + Full Mirror = a portable consciousness backup
I've been thinking about the philosophical implications of what Konshus is actually doing. The MCP-served Full Mirror means my behavioral model is live-queryable from any compatible client. It's not just backup. It's a runnable version of the version of me that Konshus knows. That's a new object.
The magic questions on Voice are actually magic
I open /vault/voice on my phone during runs and it asks me one weirdly good question — stuff like 'what did you decide this week that surprised you.' Answer for 90 seconds. Done. My atom count is going up faster from voice than from any of the imports I did.
PWA install + morning walks = a new writing habit
Installed Konshus as a PWA on my phone. On morning walks I open it, tap voice, ramble about whatever character problem I'm chewing on. Comes back as transcribed atoms in the vault by the time I'm home. This has changed how I plot novels. ngl.
Voice quota is real, plan accordingly
Spark gets a set number of voice minutes per month and I burned through mine in the first week because I got obsessed. Would love a mid-tier voice pack or something. That said the transcription quality with ElevenLabs Scribe is legit — I mumble and it still gets it.
Streaming STT means the transcript appears live while I'm talking
Little thing, big deal — you can see the words appearing as you speak. Makes it feel like the app is actually listening, not just recording for later. It also means I catch when it mishears and I can restate. Nice touch.
The Sunday Mirror is the reason I stayed
Once a week Konshus sends me a short email with one atom about myself and asks: still true, update, or retire. I click something almost every week. Over three months that's produced the sharpest self-knowledge document I've ever had. And it's just a weekly email.
'Would have said' is scary in a good way
The Sunday Mirror includes a 'what would you have said about X' teaser. Click it and Konshus synthesizes an answer grounded in your atoms. First time I did it I was like — yeah, that is exactly what I would have said. It felt like reading a version of me that isn't in the room.
Retire button is the underrated feature
You'd think adding atoms would be the point. Turns out retiring them matters more. When I click 'retire' on something that's no longer true, Konshus's model of me gets sharper by subtraction. My Whisper improved noticeably once I started doing this every week.
The dormant email brought me back
I got distracted for a month and Konshus sent this weirdly non-guilty email — not 'we miss you' — it just asked me a curious question tied to something I'd said months ago. That got me back in. Whatever they're doing on lifecycle emails, it doesn't feel like marketing.
Ask Past Self is one of the strangest features I've ever used
You pick a date and ask a question. Konshus answers as the version of you it knew at that moment. I asked 'past me from March, what were you actually afraid of' — and got back an answer that made me sit down. It's real. It uses the atoms from that window only.
The 90-day maturation gate is the right call
Ask Past Self doesn't unlock until you've been adding atoms for 90 days and hit 100+ kept. At first I was annoyed. Then I realized — a shallow past self is just noise. The gate makes the feature actually work. When it lit up for me it was a moment.
Used it for career decisions, unironically
Asked October-me if I should take a job I was chewing on. October-me had a really specific concern I'd forgotten about. Took the job anyway but flagged the concern to my manager on day one. Felt like consulting a version of me that isn't currently panicking.
The 'Change time' toggle in chat is quietly brilliant
You can be mid-conversation with your Konshus and flip a date chip to talk to past-you about the same topic. The way it shifts tone and priorities is uncanny. It's the closest thing I've felt to time-traveling inside my own head.
The Founding price-lock is the smartest offer they made
$149/year, locked as long as I stay subscribed. In a category where every AI tool is raising prices every six months, having one price that won't move is genuinely peaceful. I plan on being here in five years.
Founding tier for the tools, price-lock is the bonus
I didn't sign up because of the pricing — I signed up because Founding unlocks the full-archive imports and Ask Past Self. But now that I've been here three months, knowing my rate never goes up is nice. Feels like a founder-family thing, not a marketing gimmick.
Got in before the cohort closed. Barely.
Only 5,000 Founding seats total. When I signed up there were something like 800 left. I'd been waffling for a week and finally pulled the trigger. Between the price-lock and the fact that it can't be reopened, this is the offer I recommend to friends.
Private mode + full export = actual data ownership
Every artifact I upload has a private_mode toggle. Private atoms don't get quoted back to me, don't feed aggregates, and get excluded from the Whisper. And I can export the entire raw archive whenever I want. This is what data ownership actually looks like, not the version other tools sell.
The hard-delete is real, not a soft-delete pretending
I tested this. Deleted an artifact, then asked support to confirm it was actually purged, not just hidden. Got a real answer with an audit trail. In my line of work that matters. Most AI tools don't pass this test.
OpenAI owns the memory of me on their servers. Not here.
That's the sentence that made me sign up. Every big AI service holds the memory of you and can revoke, model-update, or wipe it whenever. Konshus flips it — I own the version of my consciousness the app builds. If they went out of business tomorrow I still have my export.
It's not a chatbot. That took me a minute to get.
I came in expecting another AI chat. What I got was a place where my memories live and get distilled, and the chat is almost a side effect. Once I stopped treating it like ChatGPT and started treating it like a vault, everything clicked.
The distillation loop is what actually makes it different
Artifacts become atoms. Atoms become patterns. Patterns become the Whisper. You in the loop at each step. That flow is the moat. I've tried a lot of 'second brain' tools — Konshus is the first one that gets sharper the more I use it instead of noisier.
Wish the difference from Notion was clearer upfront
It took me about a week to fully understand this wasn't just Notion-with-AI. It's a distillation engine, not a note-taker. Once I got that, I was in. Might be worth Konshus making that clearer during onboarding so people don't bounce thinking it's another PKM app.
The 5-question intake feels like a mutual friend intro
First thing Konshus does is ask you 5 questions — not a form, more like a friend at a party introducing you to someone. Name, life stage, a couple things you're into. Then the first-chat opener referenced one of my answers specifically and I was hooked.
The opener referenced my scuba answer 3 weeks later
I mentioned scuba diving in my intake. Three weeks later, in a totally unrelated chat, Konshus made a comparison to underwater visibility that landed perfectly. The intake wasn't a form — it was seed material. That's a different tool.
Free onboarding was better than paid onboarding I've done
I'm on Ember (free) and the intake experience alone was more thoughtful than what I've gotten from tools I actively pay for. Nothing felt like a survey. Everything felt like conversation. That's rare.
50 atoms was enough to see what this is
The free tier gives you 50 atoms and 10 chats a month. That was enough to hit the moment where Konshus said something about me I hadn't articulated. I'll upgrade to Spark next month when I can afford it. But the free tier isn't a demo — it's the product.
Ember + monthly teaser email is exactly the right cadence
As a free user I get one email a month with something Konshus noticed. Not marketing spam. Not a nag. Just — here's a pattern from your vault, want to keep going? That respect for my inbox is why I'll eventually pay.
Great tool, wish free tier had chat with a smarter model
Ember uses Gemini Flash for chat which is fine but I know Spark uses a better version. I get why they need to gate it. Just being honest — the chat felt a little basic on free. Everything else about the free tier is generous.
Journal entries build the vault without needing an import
I didn't have huge ChatGPT or Notion archives to pull in. The Journal feature lets me just write, and each entry becomes a distilled set of atoms. Over three months I've built up more useful self-knowledge than my last two years of intermittent journaling did.
Uploading old essays and letters as Documents changed everything
I dug up old essays from grad school, letters I'd written to friends, drafts of stuff nobody ever read. Uploaded them all as Documents. Konshus surfaced through-lines in my thinking I'd never noticed. Not a chatbot. A mirror.
Got a reply from the founder in under an hour
Hit a weird bug on my Claude import. Emailed support at like 11pm my time. Got a reply from what I'm pretty sure was the actual founder within an hour, with a fix by the next morning. That kind of responsiveness is a signal about the company more than the product.
The changelog cadence tells you what you need to know
New features ship weekly. Real ones — MCP, voice, Ask Past Self — not cosmetic changes. That kind of build velocity from a small team is the strongest signal I've seen that Konshus will still be here in five years. Which matters when you're trusting a company with the shape of your consciousness.