AI tool listing.
kuku is an open-source, local-first Markdown knowledge workspace for macOS. Its source of truth is a folder of ordinary `.md` files, with wikilinks, backlinks, full-text search and graph navigation layered on top rather than a proprietary cloud database.12 The same folder can remain usable with other Markdown-oriented tools, materially reducing lock-in.
The AI layer is more ambitious than a sidebar chatbot. Current product and repository material describe Agent, Ask and Inline workflows that can search the vault, read files, propose edits, create links and use selected text or attached files as context.2 Proposed changes are shown as reviewable diffs before application. That approval step is one of kuku's strongest design choices: an assistant that can alter a knowledge base is safer and easier to audit when changes are inspectable instead of silently committed.
kuku also frames accepted or rejected AI proposals as part of a longer-term “Second Brain” workflow. The repository describes decision documents and wiki updates intended to improve context inherited by future conversations.2 This is promising, although there is not yet independent evidence measuring retrieval accuracy or whether this memory design improves outcomes over mature alternatives.
The current architecture has moved beyond the product's January launch state. An older first-party build note described React/Tiptap, Gemini tool calling, no built-in cloud sync and no plugin system.3 The current repository instead describes SolidJS/ProseMirror components, encrypted sync foundations, web/server infrastructure and self-hosting support.24 Current facts therefore need to come from the live product/repository rather than launch-era reviews.
Platform availability is still narrow. The official repository lists macOS as supported and Windows and Linux as coming soon.2 The main repository contains desktop, web, Go server, protobuf, Rust AI/indexing and Docker infrastructure, and the public roadmap positions self-hosting as an alternative to kuku's managed service.24
kuku is unusually easy to evaluate today because the current public beta gives everyone full Pro-tier access with no kuku billing.5 The permanent Free tier is listed at $0/month and includes the native macOS editor, plain files and folders, wikilinks/backlinks/graph view, local indexing/full-text search and bring-your-own Gemini key for AI chat and edits.5
The planned Pro tier is $10/month and adds a managed kuku.mom account/dashboard, remote AI requests with usage tracking, encrypted sync workspaces/devices/key envelopes, managed storage quotas and conflict recovery. It is explicitly marked “Coming soon.”5 Team is listed at $15 per seat/month, also coming soon, with shared vault context, shared second-brain memory, permissions and access to major model-provider families. Enterprise is custom-priced.5
Those prices should be treated as current published intentions rather than mature commercial history. The product is still in public beta and the managed plans have not accumulated meaningful billing/support evidence. The free tier is structurally important rather than a crippled demo: local vaults have no published local cap, and BYOK AI gives users direct control over their Gemini credentials and provider spend.5
kuku's strongest trust argument is architectural. Plain Markdown remains local, the code is public, and the project allows a path from fully local use to managed sync or self-hosted infrastructure.24 The current repository describes encrypted sync using workspaces, device/key envelopes, signed commits and encrypted objects so the managed server does not need plaintext note content.2
The privacy policy is more generic than the engineering documentation and should be read separately. It says local device data remains under the user's control unless the user chooses to sync or upload it, while optional cloud/sign-in features process the data needed to provide those services.6 It also says kuku may collect technical data such as device type, app version and crash logs and usage data needed to run or improve the product, and may use infrastructure or analytics providers.6 This is less absolute than older “no telemetry without explicit opt-in” roadmap language.4
AI processing also depends on configuration. The Free plan explicitly supports bring-your-own Gemini key, while planned paid tiers add remote AI requests and broader model-provider access.5 “Local-first” should not be interpreted as “all AI inference is local.”
Independent evidence is still thin. Product Hunt currently reports 4.5/5 from two reviews and around 950 followers.7 One visible reviewer praises open-source/local-first ownership and asks for a stronger plugin/extension ecosystem; the maker agrees that extensibility is still being stabilized.8 Two reviews are useful qualitative evidence but nowhere near enough to support a statistically meaningful satisfaction score.
OpenApps' repository analysis characterizes the project as actively maintained, open-source and young, with strong recent development activity but low maturity relative to established tools.9 Repository statistics and automated “health” scores are not product-quality benchmarks, so they are useful mainly as maturity context.
There is no robust current evidence for long-term sync reliability, recovery from difficult merge conflicts, AI edit quality across large heterogeneous vaults, battery/memory behavior on large datasets, enterprise administration, or support responsiveness.
kuku is best suited to macOS users who already like plain-text knowledge systems but want AI to work directly with their files. Writers, researchers, developers and knowledge workers who use Markdown, git, wikilinks or Obsidian-style vaults can adopt it without moving their core notes into a proprietary format.12
It is particularly attractive to users who care about exit options. A local folder of Markdown files remains useful if kuku disappears, pricing changes or the user decides to switch editors. The open-source server and self-hosting direction further reduce dependency on the managed service.4
The reviewable-diff model is also a good fit for people who want AI assistance without giving an agent silent authority over their knowledge base.2
kuku is a weaker fit for users who need Windows or Linux today, a polished mobile workflow, mature real-time collaboration or a deep plugin marketplace. Product Hunt feedback already points to extensibility as an important missing layer.8 Organizations that need centralized policy controls, formal compliance documentation, enterprise support history and proven multi-user administration should treat the current Team/Enterprise roadmap as future capability rather than established deployment evidence.5
Rating: Coming soon.
Products matched by shared tasks, audience, workflow, product type and capabilities.