← Research & Data
Research & Data · Research / Knowledge Layer

Cluing

AI tool listing.

6.9OUT OF 10 · LOW CONF.

Product facts

Category
Research & Data / Research / Knowledge Layer
Pricing
See official site
Free access
Not confirmed
Platforms
Not confirmed
API
Not confirmed
Open source
Not confirmed
Integrations
Not confirmed
Last updated
2026-08-25
PRODUCT / PRICING SOURCE ↗

Cluing review

Updated 2026-08-27

Product & capabilities

What Cluing is now

Cluing is no longer adequately described as a web highlighter. The product still starts with capture: users can save webpages, passages, PDFs, Kindle highlights, YouTube material, social posts, emails/newsletters and notes into Topics and Clusters. In 2026 it has expanded into a searchable knowledge layer, AI Chat, Canvas, a Knowledge Graph, MCP connectivity, custom connectors, dynamic skills and hosted agents that can keep working after the user's computer is closed.123

Capture and provenance remain the core strength

Cluing's durable advantage is the capture layer. The current Chrome extension can save text from webpages and PDFs, import Kindle highlights, capture social and video material, receive forwarded email/newsletters and organize content into searchable Topics, Clusters and a Knowledge Graph.2

The key distinction from a generic notes app is provenance. Cluing is built around preserving the source, selected passage and user reasoning together. That is especially relevant to research-heavy work where an AI-generated conclusion is more useful if the underlying evidence remains easy to retrieve.

The product also supports collaboration: Topics can be shared, role-based access can be applied, comments can stay attached to material, and content can be exported or moved into services such as Google Docs, Notion, Gmail and Outlook.12

For researchers, analysts, consultants and legal/compliance users, that source-context architecture is more meaningful than the generic second-brain label.

AI Chat is multi-model

Cluing's AI layer is deliberately multi-model. Its current Chrome listing says AI Chat can work with frontier models from Claude, ChatGPT/OpenAI, Gemini, Grok and DeepSeek, while stored Topics can be brought directly into conversations as context.2

Cluing also exposes an editable AI Canvas and MCP integration, allowing external MCP-compatible agents to read from and write back to the workspace. This can reduce copy-paste friction and makes the knowledge base potentially useful as a shared context layer rather than a closed notes destination.2

Mobile

Cluing now has iOS and Android apps. The Apple listing identifies Collabwriting Inc. as the developer and describes capture, collaboration, Topics/Clusters, AI Chat and AI Canvas. Current version history also references Hosted AI Agents, scheduled autonomous agents, dynamic skills, custom connectors and one-click website/app publishing.4

Google Play reports more than 1,000 Android downloads and an August 18, 2026 update. It describes the same core capture and AI-workspace functionality. No representative aggregate star rating was exposed in the captured listing, so none is used numerically.5

The Android scale is modest, but mobile matters strategically because Cluing's hosted-agent proposition is partly that work can continue without a user's main computer remaining active.

Pricing & access

Pricing

Cluing's current site presents Free at $0/month with capture integrations, unlimited snippets/topics, five clusters, five guests and selected AI models. Pro is shown at $12/month when billed yearly ($144/year), with 2,500 credits per month, every frontier AI model, unlimited clusters, deep research and shared memory. The displayed Team configuration is $140/month billed yearly ($1,680/year) for three people, one agent seat and 8,000 credits per month.1

The interface also exposes configurable tier families, so these should be read as current reference configurations rather than universal prices for every seat/credit combination.

The free tier is substantial enough to evaluate the capture/knowledge architecture before paying. Pro becomes more attractive when the same stored context is repeatedly reused across several model families. Team pricing is harder to judge without hands-on evidence of how quickly real agent workloads consume credits.

Evidence & trust

Hosted Agents are the biggest change—and the biggest evidence gap

The most consequential 2026 feature is Hosted Agents. Cluing positions them as agents running on its infrastructure, remaining active after the user's computer is closed, sharing team context, using evolving skills and custom connectors, accepting tasks from teammates and publishing websites, dashboards or apps.13

Product Hunt's current launch page reports that Hosted Agents ranked #5 of the day on August 19, 2026. Maker material describes remotely continuing work, reusable connectors, encrypted credentials and workflows that can be handed between desktop, phone and teammates.3

The knowledge-management layer has several years of product history under the former Collabwriting identity. The hosted-agent layer is new enough that durability, error recovery, permission handling, connector safety and long-running-task reliability still need independent testing.

Browser-extension evidence is positive but small-sample

The Chrome Web Store provides the strongest exact-product numeric user signal currently available. The current Cluing extension is Featured, follows Google's recommended practices, has a 4.8/5 rating from 32 ratings and about 8,000 users. The listing was updated August 12, 2026.2

That is encouraging but not mass-market evidence. Thirty-two ratings support a conclusion of strong early browser sentiment, not mature-market reliability.

The Store description matches the current product direction: capture from web/PDF/social/video, team organization, AI Chat, Canvas, MCP integration and exports. The developer also declares that data is not sold to third parties and that personally identifiable information and website content are handled by the extension.2

Privacy and security

Cluing's current website and Chrome listing state ISO/IEC 27001:2022 certification. The website also describes shared connectors with encrypted secrets and hosted agents operating on Cluing infrastructure.12

The public privacy policy, however, is dated September 26, 2022. It describes account and usage data, logs, analytics/support providers, retention and GDPR/CCPA rights.6 The age of that policy is a transparency weakness because the product has changed substantially since 2022, particularly around AI routing, connectors and hosted agents.

A newer Help Center article provides a clearer deletion rule: deleting a personal Cluing account immediately deletes data from the personal workspace and removes the user's email from newsletter/support systems. For Team Workspaces, personal identifiers are anonymized while contributed content and comments can remain for team continuity.7

The Chrome Web Store additionally says the extension handles personally identifiable information and website content and carries the developer declaration that data is not sold outside approved-use cases.2

For sensitive professional material, ISO certification and explicit deletion behavior are positives. The mismatch between a rapidly evolving agent product and a 2022 public privacy-policy date remains the main disclosure concern.

Reliability and missing evidence

There is no current independent benchmark for Cluing as an application. Benchmarks such as ARC, SWE-bench, Aider, BFCL, LM Arena or Artificial Analysis evaluate models or agents under configurations that do not map cleanly to a multi-model knowledge workspace.

The evidence gap matters most around Hosted Agents. Cluing makes ambitious claims about continuous execution, publishing, connectors and shared skills. The surfaces are live, but there is no independent measurement of failure rate, task-completion quality, security isolation or credit efficiency.

Who it's for

Who should consider Cluing

Cluing is particularly relevant to researchers, analysts, consultants, lawyers, compliance teams, cybersecurity professionals and content teams that repeatedly collect evidence from the web and need exact references later.

Individual users should start with Free and evaluate whether Topics, Clusters and source capture genuinely replace part of their current browser/notes workflow. Pro is most rational when the same stored context is used regularly across several frontier models or Deep Research.

Teams should evaluate Hosted Agents separately from the mature capture layer. The architecture is promising, especially for shared context and work that should continue off-device, but buyers should run representative workflows before treating the agent layer as production infrastructure.

Strengths & weaknesses

Strengths

Cluing is strongest when work begins with external information that should remain traceable. Capturing exact source context before feeding it to AI is a better architecture for serious research than asking an ungrounded chatbot to reconstruct provenance later.

The product is also coherent across individual and team context. Saved knowledge, AI Chat, Canvas, MCP, connectors, skills and hosted agents all operate around the same workspace rather than as unrelated add-ons.

Its browser footprint is another positive. An 8,000-user extension with a 4.8/5 rating is not huge, but it is enough to show adoption beyond a private beta.2

Finally, the free tier lowers evaluation risk. Users can test whether Cluing improves their research workflow before paying for broader model access or agent credits.1

Weaknesses

The main weakness is evidence maturity. The most ambitious part of today's Cluing—the hosted autonomous-agent layer—is also the least independently validated.

The second weakness is pricing predictability for agent-heavy use. Credits are explicit, but without independent workload data it is difficult to translate 2,500 or 8,000 credits into a reliable monthly volume of complex work.

The third weakness is documentation freshness. A privacy policy last updated in 2022 predates the current multi-model, connector and hosted-agent architecture by years.6

The fourth is workflow lock-in. Cluing's value rises as users intentionally capture and organize context inside it. People with a mature existing knowledge system may face migration and habit costs.

SOURCES

Sources & references

7 sources
  1. Cluing officialCluing - The Knowledge Layer for the AI Era
    OFFICIAL
  2. Chrome Web StoreCluing - AI Knowledge Base
    USER REVIEWS
  3. Product HuntCluing
    USER REVIEWS
  4. Apple App StoreCluing App - App Store
    USER REVIEWS
  5. Google PlayCluing - Apps on Google Play
    USER REVIEWS2026-08-18
  6. Cluing officialCluing Privacy Policy
    OFFICIAL2022-09-26
  7. Cluing Help CenterWhat happens to my data if I delete my account?
    OFFICIAL2026-01-26

Related AI tools

Products matched by shared tasks, audience, workflow, product type and capabilities.

AI
CoreBase
CoreBase has matured from an early-access data assistant into a credible governed-agent infrastructure layer: live data access, per-user scoping, read-only defaults, approval-gated writes, audit trails, an embeddable agent API and an open-source outbound-only CoreMCP bridge are a coherent architecture. The main blocker is contractual maturity: current Terms contain generic anti-automation restrictions that conflict with the API/SDK product and expressly exclude important regulated-use categories even while Enterprise is marketed for regulated and air-gapped environments.
WHY RELATED · Similar input/output: Documents → Agent responses · Same category
6.5/10AITOOLMAP RATINGRATING CONFIDENCE · LOW
AI
kuku
kuku is a credible local-first Markdown workspace for macOS with unusually strong portability and reviewable AI editing. Its open-source architecture, plain-file vault, encrypted-sync direction and current free public-beta access are compelling, but the product remains young, macOS-only and supported by only a tiny independent user sample; the managed sync/AI tiers are still evolving.
WHY RELATED · Similar input/output: Documents → Notes
7.1/10AITOOLMAP RATINGRATING CONFIDENCE · LOW
AI
Unabyss
Unabyss is a coherent solution to a real multi-AI problem: it builds one continuously updated, structured context layer from work apps and exposes controlled slices to multiple AI clients over MCP. Its first-party privacy and security documentation is unusually detailed for a young AI infrastructure product, but independent exact-product evidence is still extremely thin and SOC 2 Type II remains in progress. The product is most compelling for heavy multi-agent users who understand the privacy consequences of centralizing professional context.
WHY RELATED · Similar input/output: Documents · Shared capabilities: mcp
7.9/10AITOOLMAP RATINGRATING CONFIDENCE · LOW