Cluing Review (2026): Pricing, Features & Honest Verdict

Cluing is a knowledge capture tool with hosted AI agents that read your saved research and ship work from it. We found a capable product wrapped in the worst pricing contradiction we have documented: two complete, live, incompatible price ladders on one domain. Best for: researchers and small teams. Price: From $12/mo (free plan: yes). Rating: 7/10.

What is Cluing?

Cluing calls itself the knowledge layer for the AI era, which is vaguer than what it actually does. In practice it is three things stacked. You capture snippets from anywhere you read, the tool files them into topics and clusters without being told how, and then hosted AI agents work against that pile as context: answering questions, running multi-step research, and publishing dashboards and sites to your own domain.

The capture surface is the widest we have seen in this category. Web pages, YouTube, PDFs, LinkedIn, X, Reddit, Quora, Kindle highlights, email and newsletters, plus mobile capture. One click saves the source, your highlight and your own note attached to it. That last part matters more than it sounds, because a highlight without your reaction to it is close to worthless six months later.

The company behind it is Collabwriting Inc., and Cluing is the successor to its Collabwriting product. It claims 18,253 teams and professionals. The launch team listed on Product Hunt is Sandra Idjoski, Ivan Ralic and Gordana Laskovic. It launched on Product Hunt on 19 August 2026.

The newest and most interesting piece is hosted agents. These run on Cluing’s infrastructure rather than your machine, carry your shared team context, and can publish output live with HTTPS on a custom domain. The site shows Claude Code running on Cluing and deploying a dashboard. That is a real jump from note-taking into doing work.

What are Cluing’s key features?

A knowledge base that files itself

Snippets sort themselves into topics and clusters as you save them. Clusters are shareable workspaces that group related research, and you can invite guests into one without giving them your whole library. We rate self-organising higher than most people do, because the reason research tools die is not bad capture, it is the hour of manual filing nobody ever does.

Hosted agents that publish, not just answer

Every AI model plus your own Claude Code can run on Cluing’s infrastructure, with one-click publishing to a custom domain and issued certificates. Agents run around the clock to research, monitor and organise. They are an add-on rather than an included feature, priced per agent, which we cover below. The distinction from a chatbot is that the output is a deployed artefact, a dashboard or an app, not a message.

Shared team memory and encrypted connectors

Teams share living context, dynamic skills, connectors with encrypted secrets, tasks and the agents that run them. The framing is that your whole team and your AI work from the same brain. Connectors keep credentials encrypted, and admin plans add custom roles, model access control and per-member spend limits. Per-member spend limits are a good sign in a credit-metered product, because they are the control that stops one enthusiastic colleague spending the month’s balance on a Tuesday.

Deep research with a defined cost

Deep research runs multi-step, reading across your sources and the web. Cluing publishes what it costs in credits, and the numbers reconcile: roughly 3.33 credits per AI answer and 50 credits per deep research, holding consistently across all four tiers. That sounds minor. It is not. Plenty of credit-priced tools we review never define what a credit buys, so a buyer cannot convert the price into work. Cluing can be metered before you commit.

Why does Cluing publish two different price ladders?

This is the finding that decided the score. Cluing’s pricing page and Cluing’s homepage each render a complete, internally consistent, fully rendered price ladder, and the two do not agree on anything that matters. We loaded both in a real browser to be sure we were comparing rendered pages, not stale caches.

The pricing page shows Free at $0 with 100 credits a day, Pro at $20 a month or $12 billed yearly with 4,000 credits, Superhuman at $60 or $36 with 20,000 credits, Team at $200 or $120 with 80,000 credits and 5 seats, and Enterprise as custom. It advertises a 40% saving on yearly plans.

The homepage pricing block shows Free at $0, Pro at $42 a month or $504 billed yearly with 2,500 credits, and Team at $135 a month or $1,620 yearly with 2,500 credits. Top-up credits are listed at $10 per 1,000. There is no Superhuman tier at all.

So Pro is either $20 with 4,000 credits or $42 with 2,500 credits. That is 2.1 times the price for 37.5% fewer credits, which works out at $0.005 a credit on one page and $0.0168 on the other, a 3.4-fold difference in unit cost. Team is worse: $200 for 80,000 credits is $0.0025 a credit, while $135 for 2,500 credits is $0.054, a difference of more than twenty times. A whole tier exists on one page and not the other.

The annual claim breaks too. The homepage’s $504 a year divided by twelve is $42, exactly its own monthly price, so annual billing there saves nothing, against the pricing page’s advertised 40% discount. The one figure both pages agree on is the hosted agent add-on at $25 a month or $180 a year each, and that one does carry a real 40% annual saving.

We think the homepage block is an abandoned older ladder that shipped before the current pricing page and was never taken down. That is the charitable reading and probably the correct one. It is also not the buyer’s job to work out which of a vendor’s two price lists is live. Anyone who reads the homepage, does the maths and budgets for $42 will be wrong, and anyone comparing Cluing to a rival on the homepage numbers will reject it for being three times more expensive than it is.

Can you find out what extra credits cost?

Not from the pricing page, which is the second problem. Cluing lists five one-time credit packs at 500, 1,000, 2,400, 5,000 and 10,000 credits, each applied instantly and valid for 90 days. Every one of them shows “Sign in required” where the price should be. The allowance is public. The price is not.

So the only published top-up rate anywhere on the site, $10 per 1,000 credits, lives on the same homepage block we just established is out of date. Take that rate at face value and it works out at $0.01 a credit, which is double the $0.005 a credit Pro’s own included allowance costs on the current pricing page. Buying a top-up would be the most expensive way to buy the product’s own unit.

There is also a 90-day expiry on packs. Combined with monthly credit rollover being capped at one month’s allocation on subscriptions, the message is consistent: Cluing does not want you banking credits. That is a defensible business decision, stated clearly, and we would rather have it stated than discovered.

How much does Cluing cost?

Cluing starts at $12 a month for Pro billed yearly, or $20 month to month, and there is a real free tier. We are quoting the current pricing page throughout, for the reasons above.

Free gives 100 credits a day, which is about 30 AI answers or 2 deep researches, unlimited snippets and topics, 5 clusters, 5 guests, 2MB document uploads and a curated set of faster models. Pro at $20 or $12 yearly gives 4,000 credits a month, every frontier model including GPT, Claude, Gemini, Grok and DeepSeek, unlimited clusters, 100MB uploads, 25 guests, credit rollover and unlimited relaxed answers on base models once credits run out. Superhuman at $60 or $36 gives 20,000 credits, 1GB uploads, unlimited guests, priority queues and early access. Team at $200 or $120 gives 80,000 credits shared across 5 included seats, SSO, custom roles, model access control, per-member spend limits, indemnification and an SLA. Enterprise is custom. Hosted agents cost $25 a month or $180 a year each on any paid tier.

Included credit rates descend properly as you move up: $0.005 a credit on Pro monthly, $0.003 on Superhuman, $0.0025 on Team, and $0.003, $0.0018 and $0.0015 respectively on annual billing. That is a well-built ladder.

Here is the part worth pausing on. Free’s 100 credits a day is roughly 3,000 a month, against Pro’s 4,000. Pro buys you 33% more volume. So the upgrade is not really about capacity at all: it is about every frontier model instead of a curated few, unlimited clusters, 100MB documents, rollover, and a commercial licence, because Free is personal and non-commercial use only. Buy Pro for the licence and the models. If you are choosing it for the credits, you are paying $20 for a third more of something you already get.

Against rivals, Cluing is priced fairly. Readwise, the closest capture-and-review comparison, runs $5.59 a month for Lite and $9.99 for the full product billed annually, with a 30-day trial. Recall offers a free tier with paid plans around $10 to $12 and higher tiers near $23 and $38. Cluing at $12 annual sits in that band while doing considerably more, and neither competitor offers hosted agents that deploy to your domain.

Who is Cluing best for?

Use Cluing if you read a lot across many sources and keep losing what you read. The capture range and the self-filing are the strongest parts of the product, and the free tier is generous enough to prove the habit sticks before you pay anything. Do that first.

Use it if you want research and doing in one place. The hosted agents publishing to your own domain are the real differentiator against every note tool on the market, and $25 a month for an always-on agent carrying your accumulated context is a fair price for what that is.

Use it if you are a small team that would benefit from shared context. Team at $120 a month annually for 5 seats and 80,000 shared credits is the best unit economics on the ladder, and per-member spend limits and model access control are the right controls.

Skip Cluing if you need to do commercial work on a budget of zero. The free tier is explicitly non-commercial, so client work means Pro at minimum. That is a licence question, not a features question, and it is easy to miss.

Skip it if you cannot tolerate pricing ambiguity. If you need a signed-off number for a procurement process today, a vendor publishing two contradictory ladders and hiding its top-up prices behind a login is a hard conversation with finance.

What are the best Cluing alternatives?

Readwise and Reader is the mature, boring, reliable option at $5.59 a month for Lite or $9.99 for the full product billed annually, with a 30-day free trial. It is better than Cluing at the reading and revisiting loop, has years more polish, and syncs highlights out to Notion and Obsidian. It has no agents and no team context layer. If all you want is to capture and remember, this is cheaper and does that job better.

Recall is the closest direct competitor on the AI knowledge base framing, with a free forever tier, paid plans around $10 to $12, and higher tiers near $23 and $38 for bulk actions and frontier models. It handles summarising and linking a personal knowledge graph well. It is weaker on team collaboration and does not publish anything to your domain.

Notion with AI is the default answer for teams that already live in Notion. You lose the self-organising capture and the breadth of sources, and you gain the fact that your team is already there and your documents already live in it. For many teams that beats a better tool nobody adopts.

Final verdict: is Cluing worth it?

The product deserves a better score than we are giving it. Capture breadth across web, YouTube, PDF, LinkedIn, X, Reddit, Kindle and email is best in class, the self-filing into topics and clusters solves the real reason knowledge tools get abandoned, the credit unit is defined and reconciles across every tier, included rates descend properly as you move up, and hosted agents publishing to your own domain with a shared team context layer is something no direct competitor offers at this price. Per-member spend limits and a documented rollover policy show a team thinking about how buyers actually get burned. At $12 a month annually, Pro is good value against Readwise and Recall.

The pricing presentation is what pulls this to 7. Two complete, live, contradictory price ladders on one domain is the most severe version of this problem we have documented, and the gap is not cosmetic: Team differs by more than twenty times on unit cost between the two pages, and an entire tier appears on one and not the other. The only published top-up rate sits on the stale ladder, and the current page hides all five pack prices behind a login. Delete the homepage block and this is an 8. Buy it on the free tier today, upgrade for the commercial licence and the frontier models rather than the credits, and check your invoice against the pricing page rather than the homepage.