You have Notion for docs, Obsidian for notes, n8n or Zapier for automations, OpenClaw or Hermes Agent for local-first AI. Or some subset. They don't talk to each other. Each one costs money or attention or both. And the thing they all kind of want to be — a workspace that remembers what you're working on, writes alongside you, and acts on your behalf — requires you to glue them together yourself.
We've done that glue. Enough of it to know that the right answer is to stop.
Froots is one local-first desktop app that collapses six categories into one. This piece makes the honest case for the collapse, shows what you lose (we'll be explicit), and lays out three ways to adopt it that don't require detonating your current setup.
What you're doing today, probably
A representative stack for a technical knowledge worker in 2026 looks like this:
- Notion — team docs, project wikis, the occasional database
- Obsidian — personal notes, daily journal, the graph of ideas you actually think in
- n8n or Zapier — scheduled automations; "when this happens, do that"
- OpenClaw or Hermes Agent — a local agent for CLI-ish tasks, research, long-running work
- Plus: your editor (VS Code, Cursor, Zed), your terminal, a half-dozen browser tabs
Each of these is well-made. Each one costs between $10 and $40 per month. The problem isn't any of them individually — it's that the thing you actually want is what emerges when they talk to each other, and they don't.
You want your notes to know what your automations ran. You want your agent to know what your notes say. You want to draft a routine by writing a sentence into your editor. You want the same knowledge base your research agent uses to be the knowledge base your docs are written in.
Today you write markdown in one place, trigger a Zap in a second place, ask your agent in a third place, and manually copy the results back to the first place. Froots is what happens when a single app holds all four surfaces and connects them.
The six categories, consolidated
Here's how the collapse maps, feature-by-feature. We'll be explicit about what Froots matches, what it trades, and what it leaves for the specialized tool.
Notion → the block editor + vault
| Notion | Froots |
|---|---|
| Block editor with markdown export | Block editor writing native markdown — no export step |
| Shared databases | Local vault; shared vaults on Summer 2026 roadmap |
| Team pages & permissions | Single-user in 0.1; teams tier later this year |
| Cloud-only; outage = no notes | Local-first; works offline; every save auto-committed to local git |
| Your data lives on Notion's servers | Your data is .md files in a folder you choose |
Where Notion wins today: databases, team features, the sheer polish of a billion-dollar product. Where Froots wins: your data is actually yours, and the editor has an agent living in it.
Obsidian → the vault (and then some)
| Obsidian | Froots |
|---|---|
| Markdown vault, wikilinks, backlinks | Same — directly compatible with your existing Obsidian vault |
| Graph view | Graph view |
| Plugin ecosystem of ~1,400 plugins | 60+ curated built-in skills (not plugins you install) |
| AI via third-party plugins | AI as a first-class peer pane, not a plugin |
| No persistent agent memory | Three markdown memory files the agent writes and reads every turn |
| Local-first; Sync is a paid add-on | Local-first; sync on Summer 2026 roadmap |
Obsidian wins on: plugin breadth, maturity, community templates. Froots wins on: agent-native from day one, built-in skills that are curated rather than wild-west, memory system the agent actually uses.
Compatibility note: point Froots at an existing Obsidian vault and it works. Wikilinks resolve, frontmatter reads, tags surface. You can keep both apps open on the same folder while you decide.
n8n / Zapier → skills + (soon) the routine canvas
| n8n / Zapier | Froots |
|---|---|
| Visual workflow builder | Skill composition today; visual routine canvas on the summer roadmap |
| 500–7,000 pre-built integrations | 60+ curated skills (iMessage, Gmail, Slack, Telegram, GitHub, Notion, HomeKit, etc.) |
| No LLM in the loop | LLM-native: the agent picks, chains, and narrates what it did |
| Dedicated app, separate from notes | Automations live next to the notes they operate on |
| Triggers: cron, webhook, vendor events | Triggers today: cron + event + manual; webhook on roadmap |
| Closed SaaS (mostly) | Local execution; open workspace |
Be honest: if you need 500 integrations today, Zapier beats us. Our bet is that most people use ~15 integrations repeatedly, and a curated 60 covers them with more care than a directory of 7,000 half-maintained connectors. The extra cards Zapier holds — you probably don't play most of them.
OpenClaw → local-first multi-agent system
| OpenClaw | Froots |
|---|---|
| Chat-first, channel-routed agents on Discord | Canvas-first, editor-integrated single-agent-with-personas |
| Six named agents (Clawdicus, Clipper, Dealflow, Amplifier, Sentinel, DevOps) | One agent whose persona you author via markdown files |
| Background intelligence (heartbeats, morning briefs) | Foreground workspace (active editing + on-demand agent) |
| SOUL.md prose configs | IDENTITY.md + SOUL.md + USER.md + BOOTSTRAP.md editable persona |
If you want agents routed by Discord channel that run in the background while you live elsewhere, OpenClaw is a better fit. If you want one agent that lives alongside your notes and acts when you ask, Froots is. We've written the full OpenClaw comparison here.
Hermes Agent → local runtime + learning loop
| Hermes Agent | Froots |
|---|---|
| Terminal-first + multi-platform messaging gateway | Desktop-first + editor + inbox |
MEMORY.md (2,200 char cap) + USER.md + FTS5 session search + Honcho |
context.md, decisions.md, learnings.md + persona files, continuously indexed via local vector DB |
Agent-generated skills via skill_manage |
Curated 60+ skills + skill-creator stub generator |
| Runs on 6 compute backends (local, Docker, SSH, Daytona, Singularity, Modal) | Desktop app; sync server coming |
| MIT-licensed runtime | Source-available app + AGPL-3.0 sync server |
Hermes wins on deployment range, autonomous skill generation, permissive license. Froots wins on the editor + vault + widgets + mobile (roadmap) that live around the agent. We've written the full Hermes comparison here.
Three ways to adopt Froots
You don't have to switch everything on day one. Three adoption modes work, in order of disruption:
Mode 1 — Use your existing editor, point Froots at your vault
The lowest-commitment path. Keep using VS Code, Cursor, Zed, Neovim, or Obsidian itself. Open Froots and point its workspace at the same folder your existing editor already works in.
What you get:
- Froots's agent reads, writes, and remembers against your real vault
- The memory system indexes everything you've already written
- Skills work against the current folder
- You edit wherever you were editing before; Froots is an agent on the side
What you don't change:
- Your muscle memory
- Your editor plugins
- Your team's workflow
Use this if you have a setup you love and want the agent layer without rearranging furniture.
Mode 2 — Use the Froots runtime with your own UI
A more technical path for people building on top. The Froots agent runtime exposes an MCP-compatible endpoint (coming in the 0.3 release — not shipping in 0.1) that any MCP client can attach to. Today, you can:
- Run Froots as the workspace, and point external tools at your vault folder directly
- Drive the agent programmatically via the shell by running Froots from
--headlessmode (also 0.3 roadmap)
At 1.0, the runtime becomes first-class: you'll be able to run Froots as a daemon and connect Claude Desktop, Cursor, or any MCP client to it, using the Froots vault and memory as a shared brain.
Use this if you're an agent builder, a CLI-native user, or have a setup that needs programmatic access.
Mode 3 — Just use Froots
The default. One window, one app, one vault. Editor in the middle, library on the left, agent on the right. Everything connected, no glue code, no API tokens for internal plumbing.
Most users land here within two weeks of trying modes 1 or 2. The value of consolidation shows up in small moments — writing a note and asking the agent "summarize what I decided about this last month" without swapping apps; drafting an email in the editor and having the agent send it via the Gmail skill; scheduling an automation by typing a sentence into the palette.
Use this if you've been waiting for the workspace you're trying to assemble from six apps to ship as one.
What you give up
Let's be explicit. Replacing six well-funded products with one beta app means some trades:
- Notion's database polish — our editor is markdown-first; we don't have the inline calc-engine and relation-type databases Notion is famous for. If you live in those, wait or keep Notion for that specific surface.
- Obsidian's plugin marketplace — 1,400 community plugins vs. our 60+ curated skills. We trade breadth for curation; if the plugin you love isn't one we ship, you'll miss it.
- Zapier's integration catalog — we ship what we think is well-made. If you need the long tail of SaaS integrations, Zapier still wins on raw coverage.
- Hermes's autonomous skill generation — our agent can stub new skills with
skill-creator, but it doesn't write fully-formed new procedures from experience the way Hermes does. Different bets on autonomy. - Teams features — we're single-user in 0.1. Teams tier ships with 1.0.
If any one of those is a hard requirement for your workflow, the consolidation isn't for you yet. Otherwise, the trade is what you'd expect — you lose some depth in each category, and you gain an integrated whole that's better than any single tool could be alone.
The honest summary
Froots doesn't beat Notion at being Notion, Obsidian at being Obsidian, or Zapier at being Zapier. It beats having six apps at being one app. The premise is that consolidation — one window, one workspace, one agent with context across everything — is worth more than any individual app's deepest specialty.
That's a specific bet. It's not for everyone. It's right for:
- People who value owning their data. Plain markdown, local git, no cloud lock-in.
- People who write more than they query. If your day is 60% markdown + 40% acting on it, we're your stack.
- People who want an agent that has context. Not a chat box, but a peer with memory across sessions.
- Teams who don't need teams-tier features yet. Coming in 1.0.
If that sounds like you: install the 0.1 beta. It's free during the beta, and the workspace you build this week will still be yours when 1.0 ships.