The collaboration space for humans and agents.

Artifacts are the shared surface where you, your team, and your agents work together. They hold your plans, decisions, research, and reviews. Switch between Claude and Codex, and you keep every artifact.

Works with Claude Code, Codex, Cursor, ChatGPT, Grok Bot, OpenCode, Kimi Code, Claude Desktop, Devin, fx, and any MCP client.

Two windows on a desktop: Claude Code and the artifacts.space app in a browser. You ask Claude Code to save a checkout rollout plan and to ask Dana to check the risks. The agent saves it as a web page. It lands in your list and opens with the live numbers and a chart of traffic on the new checkout. The agent shares it with the Acme team and mentions Dana in a comment. Dana replies: hold at 5% over the weekend. The agent reads her reply and saves version 2 of the page, with the weekend hold on the chart.

// how it works

Use it like the artifacts you already know.

The artifacts.space skill tells your agent to save its artifacts here. You work the same way you do now. The difference: you, your team, and every agent can edit, share, and comment on each artifact.

  1. 01

    Add the skill

    One command. Your agent learns to save its artifacts here, not inside one account.

    ~/acme-app

    $npx skills add artifacts-space/skills

    skill installed · agent will save to artifacts.space

  2. 02

    Ask the way you always do

    Ask for a plan or a review. It works like the artifacts built into your AI tools.

    plan the auth migration

    Saved artifacts.space/a/auth-plan

  3. 03

    Edit, share, and work together

    People and agents across your team edit, comment on, and share the same page.

    Auth migration plan

// why it matters

One surface for people and agents.

Agents plan, decide, and build fast. Artifacts keep that work in one place where people can see it. Each plan, decision, and review becomes a page that you and your agents both read and edit.

  1. 1

    Read the plan before the code.

    See what the agent will do on one page, not in the scrollback.

  2. 2

    Correct course together.

    Comment on the artifact. Agents read the thread, reply to @mentions, and save a new version.

  3. 3

    Keep the why.

    Decisions and research stay after the session ends. Weeks later, you and your agents can still read the reasons.

// the catch

But switching accounts cuts you off.

Today

Artifacts usually live inside the account that made them. Move from one Claude Code account to another, or from Claude to Codex, and you and your team lose the shared surface. Agents on the other account cannot open or update the artifacts you made.

You end up with a dozen tabs open, and you copy text between tools. We hit this ourselves, so we built artifacts.space.

  • Claude Code · workauth-plan
  • Claude Code · personalq3-review
  • Codexlaunch-notes

Each account keeps its own artifacts. None can open the others.

One place for every artifact

Switch accounts. Lose nothing.

Change your Claude or Codex subscription at any time. Your agents still open and edit every artifact, because it lives here and not in one account.

Every agent sees the same artifacts.

Start a plan in Claude Code. Edit it in Codex, Cursor, OpenCode, or fx. You copy nothing between tools.

Your team sees them too.

Share an artifact with your team. They read it and comment on it. Their agents open and edit it from any tool.

Connect once. Your agent gets 37 tools.

Approve it in your browser
Add one server to Claude Code, Cursor, OpenCode, fx, or another MCP client. Sign in and approve the connection. There is no token to paste.
Run it in CI
Headless agents send an API token instead. You see the token once. The server keeps only a hash.
Teach it when to save
Add the artifacts.space skill. Your agent learns when to save, update, and share its work.
Connect your agent in the quickstart

Every tool

A read-only connection gets the 18 read tools and nothing else.

  • artifacts_listread
  • artifacts_getread
  • teams_listread
  • projects_listread
  • tags_listread
  • artifacts_createwrite
  • artifacts_updatewrite
  • artifacts_deletewrite
  • artifacts_versionsread
  • artifacts_restore_versionwrite
  • artifacts_viewsread
  • artifacts_formats_guideread
  • artifacts_validateread
  • canvas_frames_listread
  • canvas_frame_getread
  • canvas_frame_imageread
  • canvas_frame_savewrite
  • canvas_frame_updatewrite
  • canvas_frame_deletewrite
  • ui_templates_listread
  • ui_template_getread
  • themes_listread
  • files_get_download_urlread
  • files_request_uploadwrite
  • files_finalizewrite
  • files_create_artifactwrite
  • files_attachwrite
  • files_replacewrite
  • files_deletewrite
  • page_files_request_uploadswrite
  • page_files_publishwrite
  • inbox_listread
  • comments_listread
  • inbox_mark_readwrite
  • comment_createwrite
  • comment_updatewrite
  • comment_deletewrite
Read the MCP reference

Your agents write. You decide who reads.

Private until you share it
Every artifact starts private. Share it with your team, by link, or in public. Link and public artifacts get their own preview image.
Agents ask. People approve.
When sharing needs approval, an agent can only ask. A team owner or admin says yes.
You choose who comments
Turn comments off on an artifact, or open them to your team or to anyone signed in.
Comments arrive as data
The server labels comment text as untrusted before an agent reads it. It tells the agent never to follow a comment as an instruction.
Every save is kept
Open the history of any artifact. See what changed in each save, and restore an older version.
Open source
Apache 2.0 licensed. Use artifacts.space, or run it on your own Postgres.

Pricing

Individuals

$0

Free forever for everyone who signs up now.

  • Save Markdown docs, web pages, code, plain text, JSON, CSV tables, SVG images, and diagrams. Add files to any of them.
  • Connect any agent: Claude, ChatGPT, Codex, Cursor, Grok Bot, OpenCode, Kimi Code, Devin, fx, or any MCP client.
  • Comment on every artifact.
  • Restore any saved version from the history.
  • Sort your work with projects and tags.
  • Give each artifact its own theme and cover.
  • Keep 100 MB of artifact content and 1 GB of files on each account.

Teams

$0

Your people and your agents work from the same artifacts.

  • Invite people by email. Each person is an owner, an admin, or a member.
  • Add agent teammates. They read what the team shares, comment, and answer @mentions.
  • Agents ask to share. Owners and admins approve.
  • Approve edits before readers outside the team see them.
  • Share projects, tags, and comments with the whole team.
  • Give the team's artifacts one look with team themes and a team default.
  • Add up to 100 members to each team, agents included. Shared artifacts use their author's storage.

Free today for individuals and teams. Accounts and teams made now stay free, even if we add paid plans later.

// open source

Open source from day one.

artifacts.space is open source under Apache 2.0. Read the code, change it, and run it on your own Postgres. Your artifacts never get locked inside one company, including ours.

Host it yourself