Roadmap

What I'm building, in the open. Here's what shipped recently, what's in flight, and what's queued up next. Vote on the polls below to nudge what comes after that.

At a glance

From here to there

Q4 2025
v1.0 launches on Windows
Editor, AI chat, BYOK, file tree
Q1 2026
Workflows + version history
15 founder templates, every save snapshotted
Apr 2026
Document suite + auto-updater (v1.0.8)
.xlsx, .docx, .pptx, .csv, .rtf, in-app updates
Apr 2026
v1.5: Memory, MCP, Voice, Ollama
Local RAG, MCP server, side-by-side AI editing, voice input, Ollama as 4th provider
Apr 2026
v1.6 + v1.7: Mac notarized + trial system
Apple-notarized .dmg, license activation, 30-day full-feature trial, opt-in telemetry
May 2026
Multimodal + PDF + RAG + long-context
Images and PDFs in chat across all 4 providers, workspace PDF index, [Compress] for long chats
May 2026
Read aloud, plugins, templates marketplace
Local Piper TTS, sandboxed plugin runtime, 4 example plugins, community templates and plugins on GitHub
May 2026
Try in browser + Spanish + German
keepance.com/try sandbox, full UI translation across 421 strings
Next
Mobile reader (iOS, then Android)
Tauri 2 mobile reader app, currently blocked on Apple Developer signup
Shipped Building now Up next
Recently shipped

In your hands today

For the full version-by-version history, see the changelog.

Live

Multimodal AI input (images in chat)

Send screenshots, charts, UI mockups, photos to the AI for visual analysis. Drag-drop, paste from clipboard, or click to upload. Routed to native multimodal APIs for Claude, OpenAI, and Gemini, plus vision-enabled Ollama models. Token-cost preview before send. Try it at keepance.com/try.

Shipped May 2026

Live

PDF chat (drop a PDF, ask questions)

Drop a PDF into chat. Native PDF support via Anthropic's API for Claude (full layout, sees figures), local text extraction via PDF.js for OpenAI, Gemini, and Ollama, with a "text extracted" mode chip so you know which path you got.

Shipped May 2026

Live

PDF RAG (workspace PDFs as searchable AI context)

Flip "Include PDFs in workspace index" in Settings, Memory and every PDF in your workspace becomes searchable context for chat. Embeddings via fastembed-rs with the multilingual e5-small ONNX model (384-dim), stored locally in LanceDB.

Shipped May 2026

Live

Long-context [Compress] for marathon chats

When a chat gets long enough to bump into the model's context window, an auto-trigger modal offers to compress older turns through a fast model. Manual [Compress] button always available. Expand back, clear, or keep going. Every compression event lands in the audit log.

Shipped May 2026

Live

Read aloud (text-to-speech)

"Read aloud" button on AI responses via bundled Piper sidecar (local TTS, no cloud, no API). Useful for hands-free workflows and accessibility.

Shipped May 2026

Live

Plugin system (sandboxed runtime + marketplace)

Plugins run in a sandboxed Web Worker with a manifest-declared permission model (workspace read/write, editor selection, AI invoke, network). Plugin contributions appear in the toolbar, sidebar, command palette, and settings. Crash-isolated, with [Restart] when something goes sideways. Four day-one plugins live in the marketplace: Word Counter, Translator, Pomodoro, Mermaid Preview. Developer docs, scaffolder (npx create-keepance-plugin), and a typed @keepance/plugin-api are live.

Shipped May 2026

Live

Templates marketplace (community workflows)

Settings, Marketplace, Templates browses, installs, updates, and uninstalls community workflow templates. Catalog at github.com/keepance/community-templates with six day-one entries, GitHub Action auto-rebuilds catalog.json on PR merge. The 15 founder templates baked into the app stay where they are.

Shipped May 2026

Live

Try in browser at keepance.com/try

Pre-seeded sample workspace, working AI chat. 5 messages or 10 minutes on the shared key, then a download nudge. Paste your own Claude key for unlimited use. Take it for a spin.

Shipped May 2026

Live

Spanish + German localization

421 UI strings translated into Espanol and Deutsch. Settings, General, Language picker, plus auto-detect from your OS locale on first launch. Manually-edited strings are protected by a lock so re-translation passes don't overwrite them.

Shipped May 2026

Live

Opt-in telemetry + privacy panel

Anonymous lifecycle events (app launches, trial starts, license activations) sent only with explicit consent. Settings, Privacy shows exactly what's sent. No content, no files, no AI prompts, no email.

Shipped v1.7.2, April 28, 2026

Live

30-day full-feature trial + license activation

Every Keepance feature unlocked for 30 days, no card required. After 30 days, AI sends and workflow runs pause until license activation; existing files stay readable forever.

Shipped v1.7.0 + v1.6.0 · April 2026

Live

Mac signed + notarized (Apple Silicon + Intel)

Apple Developer ID signed AND Apple-notarized. First launch goes through the standard Gatekeeper "Apple checked it" prompt. No right-click → Open dance.

Shipped v1.6 · April 2026

Live

Memory: local RAG over your workspace

The AI remembers what you wrote weeks ago. Local vector store under .keepance/vectors/, embeddings via fastembed-rs (384-dim e5-small ONNX model). Nothing leaves your device. Use @workspace in chat or toggle "Ask my workspace" per chat.

Shipped v1.5 · April 16, 2026

Live

MCP server: your workspace in Claude Desktop / Cursor / Zed

Keepance ships its own Model Context Protocol server as a one-click .mcpb Desktop Extension. Five tools exposed (list / read / search / write / facts) over stdio. Your workspace becomes a queryable tool inside any MCP-compatible client.

Shipped v1.5 · April 16, 2026

Live

Side-by-side AI editing

Select text, ask for a revision, watch the streaming diff land where your cursor was. Per-hunk Accept/Reject. Every accepted hunk tagged in version history with author: 'ai', the original prompt, and the provider+model. Hotkey: Ctrl+Shift+E.

Shipped v1.5 · April 16, 2026

Live

Voice input + Ollama (offline AI)

Press-to-talk voice capture via bundled Parakeet.cpp / whisper.cpp sidecar (audio never leaves the machine). Ollama as a first-class fourth AI provider, auto-detects models from your local install, $0 per call, fully offline.

Shipped v1.5 · April 16, 2026

Live

Real-time API cost meter + monthly cost dashboard

Bottom-right of chat shows $X.XX this chat / $X.XX today. Settings → AI → Monthly Cost shows cumulative spend by provider, by month. Per-provider, per-model rates updated live as tokens stream.

Shipped v1.5 · April 16, 2026

Live

Document suite (.xlsx, .docx, .pptx, .csv, .rtf)

Edit Excel, Word, PowerPoint, CSV, and RTF files natively in the workspace. Round-trip preserves formatting.

Shipped v1.0.8 · April 2026

Live

Auto-updater + signed releases

Desktop app checks for updates on launch and applies them in the background. Minisign-signed artifacts. No more chasing GitHub releases.

Shipped v1.0.8 · April 2026

In flight

Building right now

After the May 2026 mega-release, the only thing actively in flight is the dedicated mobile reader. Everything else on the previous "next" list shipped.

Building

Mobile reader (iOS, then Android)

A dedicated Tauri 2 mobile reader for browsing your Keepance workspace from iPhone, iPad, and Android. Read-only first (browse, search, tap through wiki-links), edit later. The desktop app stays where the AI work happens.

Currently blocked on Apple Developer signup. The reading workflow already works today via iCloud Drive, Dropbox, Syncthing, or Google Drive, the dedicated app is a polish layer on top.

Blocked on Apple Developer signup, then TestFlight beta

Deliberate non-goals

Won't build (and why)

Honest about what Keepance is NOT going to be. Each of these would compromise a core differentiator.

Won't build

Real-time collaboration (multi-user editing)

Local-first means single-user. If you want real-time collab, Notion does this well. If you want your data on YOUR machine, Keepance.

Won't build

Cloud sync (built-in)

"No Keepance server in the data path" is a core promise. Sync via Dropbox, iCloud, or Syncthing on your own terms. The workspace folder works the same way against any synced folder, and the mobile access guide walks through every option.

Won't build

Autonomous / agentic AI workflows

"AI proposes, user decides" is a core principle. If you want autonomous multi-step AI, use Cursor (for code) or Claude Code (for code) or ChatGPT Operator (for browsing). Keepance is for workspaces where every change is reviewed before it lands.

Won't build

Subscription tier ($X/month)

Pricing is locked at $0 trial, $49 Pro, $99 Lifetime, $29 Founder's Launch (first 100 buyers). One-time pricing is a pillar. Adding a subscription tier dilutes the message and creates pricing complexity. Re-evaluable in year 2 if the math forces it, but the default is no.

Won't build

Vertical SKUs (legal, medical, academic editions)

Premature. The product is general-purpose for now. If buyer signal ever demands a vertical, I'll revisit.

You decide

What should I build next?

Vote for whatever you'd actually use. Highest-vote items get prioritized within each release window.

What should I build after the mobile reader?

The May 2026 release shipped multimodal, PDF chat, RAG, [Compress], read-aloud, plugins, templates marketplace, browser demo, and Spanish + German. Mobile reader is in flight. After that, here's what's on the candidate list. Pick one. You can change your vote anytime.

Which new workflow template would you use most?

Workflows are the structured AI interviews that produce a finished doc. The 15 baked-in templates cover most founder needs; this is what to add next.

Don't see what you need?

Email me a one-liner about what'd make Keepance a daily-driver for you. I read every one.

[email protected]

Try Keepance yourself

Free download. Pro is $49 one-time, Lifetime $99. The first 100 buyers get Lifetime at $29.

Get Keepance