Frequently Asked Questions

Keepance is a desktop app that combines a Markdown editor with an AI chat interface. Every AI conversation produces real Markdown files in a real folder on your computer. You can paste an image into chat and the AI will look at it. You can drop a PDF in and ask questions about it. You can have the AI read its replies aloud. You can install community plugins to extend the editor. It's built for indie founders who use AI to plan, write, and run their businesses, but the same shape works for anyone who'd rather keep their thinking on their own machine.

vs Notion: Notion is cloud-only, your data lives on Notion's servers and you pay them for the privilege. Notion's AI features are bolted on top, not native to the document model. Keepance is local-first (your files live on your hard drive) and AI is the primary input method, not an add-on.

vs Obsidian: Obsidian is also local-first, but its AI features come from community plugins that are inconsistent and often charge separately. Keepance has AI native, with four providers built in (Claude, OpenAI, Gemini, Ollama), a workflow template system, multimodal input (images + PDFs), local RAG over your files, and its own sandboxed plugin runtime if you actually want to extend the editor.

vs ChatGPT: ChatGPT throws conversations away. Keepance persists every conversation as editable Markdown files you own, and lets you bring whichever model you actually like.

It means your data lives on your hard drive in a folder you choose, in plain Markdown files you can read with any text editor. Nothing is uploaded to our servers. The app works offline (except when you're actually making AI calls). If we shut down tomorrow, your files are still on your computer, you just can't get updates anymore.

Bring Your Own Key. You sign up directly with Anthropic, OpenAI, Google, or run Ollama locally for free, get an API key (or just point Keepance at localhost), and paste it into the app. AI requests then go from your computer directly to the provider, Keepance is never in the middle. See the API Keys Guide for details.

No. If you can install a desktop app, paste an API key into a settings screen, and type into a chat box, you have all the skills you need. The hardest part is signing up for an AI provider account, which takes about 5 minutes and is well-documented.

Free tier: $0 forever (limited to 1 AI provider, 3 templates, 1 workspace).
Pro: $49 one-time (all AI providers, all templates, unlimited workspaces, 1 year of updates).
Lifetime: $99 one-time (everything, plus updates forever).
Founder's Launch: $29 lifetime, first 100 buyers only.

Plus whatever you pay your AI provider directly for your API usage (typically $5-$20/month for active use).

Because subscriptions feel wrong for software with no server. We don't host your data, we don't run AI inference for you, we don't have ongoing infrastructure costs that justify monthly fees. Pay once, own the version you bought, get updates if you bought a tier that includes them.

Yes, your single-user license works on multiple devices that you personally own (e.g., your desktop and your laptop), as long as it's not in active use by more than one person at a time.

Yes. 14-day no-questions-asked refund. Email [email protected] or use LemonSqueezy's customer portal.

The Software keeps working. You just don't get new updates unless you buy another year (or upgrade to Lifetime). There's no remote kill switch, if you bought v1.5, you keep v1.5 forever.

No. Your files live on your hard drive in plain Markdown format. We have no server that stores them. The only thing our server sees is your license key when you activate the app.

No. AI requests go directly from your computer to your chosen provider (Anthropic, OpenAI, or Google) using your API key. We are never in the loop. Read the Privacy Policy for the full breakdown.

No. Because everything is plain Markdown in a regular folder on your hard drive, you can keep using your files in any other Markdown editor (Obsidian, Notion, VS Code, even Notepad) forever. The Software might stop getting updates, but your data is yours.

Most people should start with Claude (Anthropic). It's the best at long-form writing and following complex instructions, which is what most Keepance workflows need. New Anthropic accounts get $5 in free credits, enough for ~50 workflow runs.

Yes. Each AI chat in Keepance has a per-chat model selector. You can have a Claude conversation about your business plan and a GPT-4 conversation about your pricing strategy in the same workspace, and switch between them whenever you want.

You pay your AI provider directly. For Claude Sonnet (the default model), expect about $0.05-$0.20 per workflow run. Heavy daily use is typically $5-$15/month. Gemini Flash has a generous free tier, 1500 requests per day, enough for most users to never hit a paid charge.

ChatGPT consumer subscriptions ($20/mo for Plus, $200/mo for Pro) are entirely separate from OpenAI's API. They're different products with different billing. To use OpenAI's models with Keepance, you need an API account at platform.openai.com, your ChatGPT Plus subscription does NOT transfer.

Yes. Paste a screenshot, drag a photo into chat, or click the paperclip and pick a file. Claude, OpenAI, and Gemini all see the image natively. Ollama sees it too if you've pulled a vision-capable local model. There's a small warning if you switch to a model that doesn't do vision, plus a token-cost preview before you send.

Yes, two ways. Drop a PDF into chat and the AI reads it for that conversation. Claude and Claude Opus get the native PDF (full layout, figures included), the other providers get the text via PDF.js with a "text extracted" chip so you know the difference.

Or flip "Include PDFs in workspace index" in Settings -> Memory and every PDF in your workspace becomes searchable context for chat. The local RAG index uses fastembed-rs with the multilingual e5-small model, runs on your machine, and never leaves the device.

Not anymore. When a chat gets long enough that you'd hit the model's context limit, Keepance pops a [Compress] modal and offers to summarize older turns through a fast model. You can also trigger compression manually, expand it back if you need the original turns, or clear the compressed summary entirely. Every compress event is logged to the audit log.

Yes. Click "Read aloud" on any AI message. Audio is synthesized by a bundled Piper sidecar that runs locally, so no audio data is sent to the cloud. Useful when I'm pacing around the kitchen and don't want to read 800 words off a screen.

Yes. There's a sandboxed Web Worker plugin runtime with a manifest-declared permission model (six permissions covering workspace read/write, editor selection, AI invocation, and network). Plugins can add toolbar buttons, sidebar panels, command-palette actions, and editor commands. Four plugins ship in the day-one marketplace: Word Counter, Translator, Pomodoro, and Mermaid Preview. If a plugin crashes, it's isolated, you click [Restart] and keep working.

If you want to write your own, see the developer docs at /docs/plugins/. There's a npx create-keepance-plugin <name> scaffolder, a typed @keepance/plugin-api package, and seven docs pages covering manifest, permissions, API, publishing, examples.

Open Settings -> Marketplace -> Templates. Browse what's there, click install, done. The catalog lives at github.com/keepance/community-templates and rebuilds automatically when someone merges a PR. Six day-one templates are already in there. The 15 founder templates baked into the app aren't going anywhere; this is the "you-and-the-community" layer on top.

Yes. Both marketplaces are open submission via GitHub PR. Submission docs at /docs/marketplace-submissions.

Yes. keepance.com/try loads a pre-seeded sample workspace and a working AI chat. You get either 5 messages or 10 minutes on the shared key, then the exit modal nudges you toward downloading. If you'd rather use unlimited messages without downloading, paste your own Claude key (BYOK) into the input and it'll run on your key for as long as you want.

English, Spanish, and German. The whole UI (421 strings, settings, marketplace, every dialog) is translated. Pick the language in Settings -> General -> Language, or let Keepance auto-detect from your OS locale on first launch. Any human edits to a translation are protected by a lock so re-translation passes won't overwrite them.

Officially supported as of v1.6: Windows 10/11 and macOS (both Apple Silicon and Intel). Linux builds (AppImage, .deb, .rpm) are produced by our CI and published on the releases page, but Linux is not officially supported in v1.6, we'll polish and announce Linux as a v1.7 release after the v1.6 launch settles.

The Windows installer is signed via Azure Trusted Signing with a Public Trust cert (publisher reads "Jameson Daines"). The warning you may see is SmartScreen's "unrecognized app" reputation warning, not a malware warning, because the cert is new and SmartScreen reputation hasn't built yet. Click "More info" then "Run anyway" to proceed. Windows remembers the choice per user. The warning clears automatically once a few hundred people install it, typically 2 to 4 weeks. You can verify the source at github.com/keepance/keepance.

v1.6 adds a portable Windows build. It's a single .exe you can drop anywhere, no install step, no Start Menu entry, no Admin rights needed. Useful if you want to run Keepance from a USB drive, try it without committing to a full install, or install on a machine where you can't run installers.

Caveats: the portable build still saves your workspaces and settings to %APPDATA%\Keepance (not next to the .exe). The auto-updater is disabled; you re-download manually. The MCP extension (.mcpb) isn't supported in portable mode. Use the full installer if you need MCP or automatic updates.

Yes, by putting your workspace folder inside Dropbox, iCloud Drive, OneDrive, or any other file sync service. Keepance writes plain Markdown files, so any file sync tool can handle it. We don't have built-in sync because we don't want to be in the cloud-storage business, and using a dedicated sync tool is usually better than what we'd build anyway.

Yes, today, as a reader. Point your Keepance workspace at a folder inside iCloud Drive, Dropbox, Syncthing, or Google Drive on your computer, then open the same folder in your phone's Files / Dropbox / Drive app. Your notes show up on the phone in seconds. See the Mobile access guide for the four step-by-step setups, including which one to pick based on what you already use.

The full source code is visible at github.com/keepance/keepance, but Keepance is closed-source proprietary software, you cannot redistribute, fork, or build a competing product from it. The source is visible so you can verify what the app does (privacy, security, etc.) but not so you can ship it yourself.

Either by email to [email protected] or by opening an issue at github.com/keepance/keepance/issues.

A dedicated Keepance mobile reader is in beta and will land in the v2.0 cycle. It'll focus on read access first (browsing, searching, and tapping through wiki-links on the phone), with editing as a follow-up once we've gotten read right. Until that ships, the supported way to access your workspace on a phone is the cloud sync workaround documented in the Mobile access guide: point Keepance at a folder inside iCloud Drive, Dropbox, Syncthing, or Google Drive, then open the same folder on your phone in the matching app. It works today on every iPhone and Android with no extra install.

Possibly, but as a separate product, not a feature added to the v1. Real-time collaboration, multi-user permissions, and team accounts are a different problem with different infrastructure and a different price point. v1 is solo-founder-first and stays that way.