weave documentation

weave vs Airtable — an open-source, self-hosted alternative

weave is a local-first, MIT-licensed work platform: connected tables, relations, workflows, formulas, rollups, and markdown documents on every record, in a single SQLite file you own. This page maps Airtable’s concepts onto weave’s, and is honest about what is missing.

Airtable’s product and pricing as described here reflect August 2026 — check airtable.com for current terms.

Concept mapping

Airtable	weave	Notes
Workspace	Workspace	One `.db` file. A server can host several side by side, switched from the icon rail.
Base	Space	Spaces group tables. A workspace holds many.
Table	Table	Qualified as `Space/Table` when names collide.
Record	Entity	Every entity gets a per-table public id — `Task#12` — usable as a ref anywhere.
Field	Field	See the field-type table below.
Grid view	Table view	Inline editing on every cell.
Kanban view	Board view	Columns are workflow states.
Gallery / list	List view
Linked record	Relation	Bidirectional and real — the inverse field is created and maintained for you.
Lookup	Lookup	Pull a field across a relation.
Rollup	Rollup	Aggregate across a relation.
Formula	Formula	Computed from other fields in the row.
Long text / rich text	Document field	Full markdown, and any number per entity — not one notes field.
Attachments	File attachments	Stored in a `files/` directory beside the workspace.
Automations	Automations	Trigger on created / field changed / state changed → set field, append doc, add comment, POST a webhook.
Interfaces	—	Not built.
Forms	—	Not built.
Sync / integrations	Webhooks + REST + CSV	Outbound webhooks and a full REST API; no prebuilt connector catalog.
Collaborators & permissions	—	Not built. weave has no accounts at all — see below.

Field types

weave implements text, number, date, date range, checkbox, url, email, select, multiselect, workflow state, relation, lookup, rollup, formula, and document.

Airtable types with no weave equivalent today: rating, duration, barcode, button, user, currency and percent as distinct types (use number), and the AI-generated field types.

Where weave wins

You own the file. A workspace is one SQLite file on your disk. Open it with sqlite3, back it up with cp, put it next to the project it describes. No export button to be held hostage by, and no row caps other than your disk.
No per-seat bill. It is a repository. Clone it.
Documents are first class. Every entity carries as many markdown documents as you want, each addressable as a URL: /e/Task#12/doc.md, .html, .pdf. Mermaid diagrams render in place. Airtable’s long-text field is not this.
Agents are first-class users. A built-in MCP server exposes 23 tools — including schema design, not just record CRUD — alongside the REST API and a scriptable CLI, all over the same engine. An agent can create the space, the tables, the relations, and the automations, then fill them.
It runs offline, with no telemetry and no account.
Nothing to install. No Docker, no Postgres, no npm install, no build step. Node ≥ 22.16 and git clone.

Where Airtable wins

Permissions and collaboration. weave has no authentication and no per-user permissions at all. Everyone who reaches the port is an admin of every workspace. That is a design position, not a roadmap gap you can wait out today.
Interfaces and forms. Airtable’s app-builder and form surfaces have no counterpart in weave.
The marketplace. Hundreds of prebuilt integrations and extensions versus outbound webhooks and an API you wire up yourself.
Maturity. Airtable has been hardened by a decade of production use across a huge install base. weave is young, and backed by its test suite rather than by that mileage.
Mobile apps, realtime multiplayer cursors, revision history at record level.

Migrating

There is no one-click importer. The practical path:

Export each Airtable table as CSV.
Create the space and tables in weave — via the UI, the CLI (table create, field add, relation add), or by handing an agent the MCP server and the CSV headers.
node bin/weave.js import or the weave_import_csv MCP tool loads the rows.
Recreate linked records as relations, then lookups, rollups, and formulas on top.

Migrating off weave is the part worth checking before you commit: node bin/weave.js export --data <file> writes the whole workspace as readable JSON, csv <table> dumps any table, and the .db is standard SQLite.