Logseq is a free, open-source knowledge outliner. Every note is a bullet in a tree, every bullet is a referenceable block, and daily notes are the front door. Around that core it has grown block references, queries, flashcards, PDF annotation, a plugin marketplace, and a Whiteboards mode — on macOS, Windows, Linux, iOS, and Android, licensed AGPL-3.0. On July 13, 2026 it shipped its long-promised 2.0 “DB version,” which moves the canonical store from plain files to a local SQLite database.
Acres is a local-first visual note app for macOS and Windows. Notes are plain CommonMark files with YAML frontmatter, boards are readable JSON, and both live in a folder you pick. The primary surface is not an outline but an infinite canvas — cards, connectors, section frames, mindmaps — and the app is a one-time purchase — a 14-day trial, then $59 once — with no account and no telemetry.
These two share more philosophy than almost any other pair we compare. Both are local-first. Both work offline. Both treat notes as property, not tenancy. Two differences matter: the primary surface — Logseq thinks in indented lists, Acres thinks in space — and, since July 2026, what “your files” physically means. Logseq’s actively developed version keeps notes in a database you export from. Acres keeps notes as files you could still edit in vi if Acres vanished tomorrow.
What happened to Logseq’s plain files?
The short version: the rewrite shipped, and it changed the deal.
Logseq officially announced the database rewrite in April 2024, promising better performance on large graphs and a foundation for real-time sync. The wait was long — long enough that “Logseq alternative” became its own genre of blog post — but on July 13, 2026 it arrived as Logseq 2.0. Development since has been active, with updates landing regularly through the summer.
What shipped is a different architecture, not just a faster one. In the 2.0 DB version, a local SQLite database is the source of truth; you no longer edit Markdown files by hand, you export to them, and Logseq’s own documentation notes that the standard-Markdown export “cannot capture all data.” The file-based app lives on as “Logseq OG” in its own repository, in maintenance mode: security patches and compatibility updates, no new features. The new real-time sync (RTC) is in alpha, will be paid as part of Logseq Pro, and is invite-only for now — backers at $5/month and sponsors at $15/month get test access. Whiteboards were temporarily removed from the DB version while the team decides how to handle them; they still work in OG. Logseq’s docs recommend regular backups of DB graphs because data loss is still possible.
The years in between cost real trust. In the Hacker News thread on the 2.0 release, long-time users describe paying for sync that took years to arrive, and several say they left for Obsidian or org-roam because plain-text storage was, in one commenter’s words, “a non-negotiable.” None of this makes the DB version wrong — real-time collaboration over a folder of Markdown files is genuinely hard, and the rewrite is what made it possible. But it splits Logseq’s original promise in two. As of August 2026, a Logseq user chooses between the version that keeps plain files and the version that gets new features. Acres exists so that isn’t a choice: the files are the product; the app is a view over them.
Is Logseq’s Markdown the same as Acres’s Markdown?
No, and the difference shows up the moment another tool reads your notes. Logseq is an outliner, so every line it writes to disk is a list item — a long-standing GitHub issue (#3457) tracks how the leading - on every block breaks the way other Markdown tools render the files. Properties are written as key:: value lines, block references as ((uuid)), queries as {{query}} — all Logseq-only syntax that arrives in Obsidian or a text editor as literal characters. A forum thread asking for “a more standard flavor of markdown” has been open since the project’s first year.
Acres writes CommonMark. A card is a document — headings, paragraphs, GFM tables, checkboxes — with properties in YAML frontmatter, where every Markdown tool expects them. Even wikilinks are stored as ordinary links ([label](acres://card/…)), so a card opened in VS Code or iA Writer reads as prose, not as a data format. Boards are pretty-printed JSON you can diff and script against. Nothing to export, because nothing was ever locked in.
To be fair, Logseq OG’s files are valid Markdown and genuinely portable — outliner-flavored, but yours. The contrast is between Markdown as an interchange format and Markdown as the native tongue.
Outliner or canvas — which fits how you think?
This is the honest fork in the road, and it has little to do with file formats.
Logseq’s outliner is the real thing. A block reference lets one bullet appear in many places and update everywhere. Queries collect blocks by tag, property, or date into living lists. Daily notes make the journal the default capture surface, and years of workflows are built on that trio. Acres doesn’t attempt any of it at the block level: links, backlinks, and filters all operate on whole cards.
What Acres offers instead is space. The canvas is the app, not a mode — cards, labeled connectors, tinted section frames, mindmaps, stickies, PDFs, and web embeds on one surface, with a measured 60fps pan on a production-scale imported board. The nearest thing to a block reference is that one card can sit on many boards: content lives once, placements live per board. Acres has a journal too — a page for every day, where entries are ordinary cards you can place on any board — and a Database tab that filters, sorts, and groups cards over shared properties with saved views: a spreadsheet-shaped answer to some of what Logseq queries do, without a query language.
Logseq does have a canvas — Whiteboards, free in OG — and if you only occasionally need to arrange notes spatially, it may be enough. But it is a side feature, currently absent from the actively developed version. If the board is where your thinking happens, a canvas-first tool is a different experience; that’s the same argument we make in Acres vs Obsidian Canvas and Acres vs Heptabase.
What about AI and agents?
Logseq ships no AI of its own; community plugins add chat integrations, and per Logseq’s DB-version docs, 65+ plugins already support DB graphs. There is no official MCP or agent surface.
Acres treats agents as first-class users of the vault. AI is bring-your-own-key from seven providers — or Ollama running entirely on your machine, which should land well with anyone who chose Logseq for privacy. The in-app agent mode (18 tools) makes destructive edits as proposals with word-level diffs, never silent writes. Beyond the panel, a bundled MCP server (29 tools) connects Claude Code, Cursor, and other clients in one click, every vault self-describes with a generated AGENTS.md and knowledge graph, and an activity ledger attributes every agent change with word-level diffs and per-change revert. Not a chatbot about your notes — a collaborator in them. The full story is on the agents page.
Who should pick Logseq?
Pick Logseq if you think in outlines — if block references and queries are how your notes earn their keep. Pick it if you need Linux or a phone in your pocket today, or if your notes live in Org-mode, which almost no other graphical tool speaks. And if open source is a requirement rather than a preference, AGPL-3.0 means no acquisition, pivot, or pricing decision can take the app from you. In all four cases Logseq is the better tool.
Pick Acres if you’re a plain-file believer who thinks spatially and wants agents working in the vault rather than a plugin talking about it — accepting that Acres is desktop-only, from a company you’d be trusting earlier in its life.
Switching
There is no Logseq importer in Acres. But a Logseq file graph and an Acres vault are both folders of Markdown, so the move is mostly mundane.
From a file graph (Logseq OG), copy your pages/ and journals/ Markdown files into an Acres vault’s cards/ folder — Acres adopts foreign Markdown live, byte-preserving the body and deriving the title from frontmatter, first heading, or filename. Your text, headings, and code blocks survive as written.
Be clear-eyed about what doesn’t carry. Pages arrive as the nested bullet lists they are on disk — Acres won’t flatten Logseq’s outline into prose. Block references arrive as literal ((uuid)) text, queries don’t run, key:: value property lines stay text rather than becoming Acres properties, and [[page links]] come across as characters, not live links. Org-mode files don’t become cards at all. Whiteboards stay behind — there’s no conversion for Logseq’s canvas files.
From a 2.0 DB graph, run Logseq’s “Export as standard Markdown” first, then bring those files over — knowing Logseq’s docs say that export can’t capture everything, so export early rather than late. Either way, once the files are in the vault they’re ordinary Acres cards: they go on boards, show up in the Database tab, and are visible to your agents. And they stay files. On disk. Sync them however you like — git, Syncthing, iCloud Drive — or use the optional end-to-end encrypted sync service.