Skip to content

Own your

Connect projects, documents and AI agents in a workspace your team can own.

Start with weave See the connections

Free to self-host · $5 cloud · Export anytime

  • AgentsMCP, REST and CLI
  • DocumentsA page in every row
  • TablesRelations and rollups
  • One SQLite fileYours to move
Example workspace

Own your context.
Manage it in weave.

Run the MIT-licensed code on the team's server, or use cloud for $5 a month. Export the workspace and import it on either plan.

Cloud or self-hosted

Getting started

Cloud or self-hosted. Start with a table and its first row.

Start a workspace
  1. 1

    First table

    Projects, customers or tasks.

  2. 2

    Linked records

    Add relations, notes and status.

  3. 3

    Shared workspace

    The team and its AI agents update the same records.

Bring everything together.

Connected tables and documents for teams and their AI agents.

  • Workspace, space, table and row: one kind of entity at every level
  • ⌘K search across workspaces on a SQLite FTS5 index, with copyable permalinks
  • tasks_final_v3.xlsx
  • Launch plan, final
  • handoff.md
  • #launch thread
  • Q4 board
PROJECT
Autumn launchAL
  • Customer interviews
  • Launch checklist
  • Launch brief
  • Thread, 14 messages
  • Agent updateSynced
Example workspace · five scattered places, one record

Connected tables

Link tasks to a project and track their status.

  • Bidirectional relations: weave creates and maintains the inverse field
  • Lookups, rollups and formulas across relations
Projects12 rows · 4 relations
Tasks148 rows
People9 rows
Documents214 pages
Customers31 rows
Your agentsMCP · REST · CLI
many to oneownerbriefaccount
Example workspace · tables, relations and the agents that read them

Row documents

Keep the brief and decisions on the task itself.

  • Any number of markdown documents per entity
  • Each document addressable as /e/Task#12/doc.md, .html or .pdf
Tasks5 rows
NameStatusOwner
Customer interviewsDoneJo Lin
Launch checklistIn progressAlex Lee
Website refreshPlannedMia Kent
Pricing pagePlannedAlex Lee
Partner emailDoneJo Lin
Tasks / Launch checklistLaunch checklist
In progressAlex LeeAutumn launch
  • Confirm launch date
  • Draft announcement
  • Brief the support team
Example workspace · every row opens as a document

Agent updates

Connect an external AI agent to update the team’s workspace.

  • MCP server built in, 64 tools
  • Undo for every entity mutation: weave undo, POST /api/undo or weave_undo

Add a Budget to Projects that totals the cost of each project's tasks.

Your agent · weave MCP
1 field added · 12 rows filled · undo any time
ProjectsSchema updated
NameStatusTasks+ Budget
Autumn launchIn progress5$12,400
Customer pilotPlanned3$3,150
Website refreshDone4$8,900
Budget = rollup · sum of Tasks.Cost
Example workspace · one sentence, one schema change

Shared workspace

The team and its AI agents update the same records.

  • CLI, server and MCP on one file, concurrently
  • Three roles: observer reads and comments, editor writes rows, architect changes structure
PROJECT
Autumn launch
3 linked tasks
DOCUMENT
Launch brief
Attached to Autumn launch
PERSON
ALAlex Lee
Owns 2 tasks
AGENT
External AI agent
Updated Launch checklist
Example workspace · the two strands of the weave mark, holding four records

Feature comparison

Connected tables, documents and agent access.

Swipe to compare all four tools.

FeatureweaveAirtableNotionClickUp
Open source, MIT
Self-hosted
One SQLite file, no database server
Relations, lookups, rollups, formulas
REST API
MCP server
Command-line interface
Bring your own AI model
No lock-in: export, then self-host

Airtable, Notion and ClickUp as published on their own sites and developer docs, October 2026.

Cloud or self-hosted

The same code on either plan.

Self-hosted

The MIT-licensed repository, on the team's own server.

Free MIT licensed

View on GitHub
  • No open-core split: every feature ships in the repository
  • Deploy files for Docker, Railway and Fly.io
  • OIDC sign-in and agent tokens
  • Full test suite and CI workflows

Cloud

Hosted weave at weave.grunion.ai. Same code as the repository.

$5 per month

Sign up
  • Use weave in the browser at weave.grunion.ai, with no server to run.
  • No built-in AI: bring your own model
  • Full workspace export at any time
  • Return to self-hosted with weave import
Local install & agent connection

Local install

git clone https://github.com/grunion-ai/weave
cd weave
node bin/weave.js serve --port 4400 --data ./my-workspace.db

Open http://127.0.0.1:4400; the server binds 127.0.0.1 and runs on Node 22.16 or later with nothing installed from npm and no build step.

Agent over MCP

claude mcp add weave -- node /path/to/weave/bin/weave.js mcp --data /path/to/my-workspace.db

Claude Code gets the 17 default tools.

Container

docker compose up -d

The container binds 0.0.0.0 for a platform proxy; the repository also ships fly.toml, railway.json and wrangler.toml.example.

MCP

claude mcp add weave -- node /path/to/weave/bin/weave.js mcp --data /path/to/my-workspace.db

Claude Code and six other clients connect to a local instance over stdio, or to weave.grunion.ai/mcp over HTTP. Sign-in happens in the browser through the OIDC provider; there is no token to paste.

REST

GET /api/tables/:ref/stats

Scripts and services call the same endpoints the web UI uses, stats and undo included.

CLI

weave build spec.json --dry-run

The 98 subcommands include build, query, backup and undo.

Release cadence.

GitHub releases per week.

Release history
2
Aug 31
11
Sep 7
11
Sep 14
18
Sep 21
13
Sep 28
25
Oct 5
17
Oct 12
2
Oct 19
Read from GitHub at build time ·

Questions.

Is weave an Airtable alternative?

Yes for linked records, rollups, formulas and views, with markdown documents on every row and an MCP server added. No for Airtable's collaborator permissions, hosted forms and marketplace apps: weave's permissions stop at three roles per workspace, and its board, calendar and timeline views are not built.

Is weave a Fibery alternative?

That is the closest description of weave: connected tables across spaces, documents attached to entities, automations and a relation graph, in a single file the team owns. docs/PARITY.md holds the feature matrix.

How close is weave to Fibery?

weave scores 81.5% against the feature list of Fibery, the product it is closest to, over a target of 80%. The score comes from docs/PARITY.md, taken against v0.4.63 on 2026-10-03, and the matrix there names every feature. Partial today: assignees and people, avatars and icons on entities, granular permissions. Not built: board and kanban view, list view, calendar view, timeline and gantt, whiteboards, reports and charts.

How does a team migrate off weave?

weave export writes the full workspace as JSON on either plan, and weave import reads it back on either.

What happens before require-auth runs?

Until an architect runs weave workspace require-auth, anyone who reaches the port can do everything an architect can; a local install binds 127.0.0.1. Editor access also needs care once auth is on: a document of kind html renders on weave's own origin, so a script an editor puts in one runs for whoever opens it. Grant editor access the way the team grants write access to a repository.

Which AI model does weave use?

None. Agents connect from outside over MCP, REST or the CLI, each with the model and provider the team chooses; weave itself makes no model calls.

Is weave open source?

Yes. The full repository is MIT-licensed, including the code used by the cloud plan.

Start a workspace.

Start with weave