Learn Artifactry in an afternoon.

1. What Artifactry is, and why it exists

When you ask an AI to write a report, build a web page or draft a deck, the result usually lives inside that one chat. Close the chat, switch to a different AI, or come back next week, and it's gone or buried.

Artifactry is a library for those results. Think of it as a shared, organised filing cabinet that both you and your AI tools can open. An artifact is one saved file (a PDF, a Word document, a slide deck, a web page or an image) together with everything Artifactry knows about it: its title, tags, versions, who can see it and how it's been used.

Three things make it different from a normal file drive:

• AI apps can use it directly. Claude, ChatGPT, Gemini and other assistants connect over MCP and can search, save and share for you.
• Links never break. Each artifact has one permanent address that survives renames, moves and new versions.
• You stay in control. An AI can only do what you allow, can never see more than you can, and has to ask before sharing anything outside your organization.

2. Getting started: create your account

Artifactry has no passwords. You sign in with a short code sent to your email, and the same steps create your account the first time.

1. Open app.artifactry.app in any browser, on your computer or phone.
2. Type your email address and choose Send code.
3. Open the email from Artifactry and type the 6-digit code on the next screen.
4. You're in. A personal workspace is created for you automatically, and you are its admin.

A quick tour

• Library lists everything you can open, newest first. Filter by type and change the sort.
• Search looks inside your files as well as their names.
• Add uploads a file or connects an AI app.
• Profile holds your connected AI apps, tags, active shares and the in-app guide.
• Workspace (the icon in the left rail on desktop) manages the current workspace's name, logo, members and invitations.

3. Your first artifact

1. Choose Add, then pick a file from your device.
2. Give it a clear title and, if you like, a short description. Pick the workspace it belongs in.
3. Choose Upload. Artifactry checks the file's real type, records a checksum, scans it, and pulls out its text so it's searchable.
4. Open the artifact. You'll see suggested tags based on its content: accept the good ones, delete the rest, or add your own.

To update it later, upload a new file to the same artifact rather than a new one. That creates version 2, keeps version 1 downloadable, and the link stays the same.

4. Workspaces and teams (share mode)

A workspace is a separate space with its own artifacts, tags and members, a bit like a project folder with its own guest list. Switching workspaces never mixes two projects together.

Turning on team mode

1. Open Workspace and go to Invitations.
2. Enter a teammate's email and choose a role.
3. The first invitation from a personal workspace asks you to confirm it becomes a shared workspace. Confirm, and the invite is sent.

Teams: reusable audiences

A team is a named group such as "Board prep" or "Finance reviews". Share with the name instead of typing emails each time. Membership stays live: someone added to the team later gets access to what was already shared with it, and removing someone takes it away.

5. Sharing

Nothing is shared by default. When you share, you choose who and what they may do.

For every share you can set an expiry, allow view only or download, and revoke it at any time, which ends access immediately. Recipients see only the item you shared, never the rest of your library.

6. Connecting AI apps

AI apps reach Artifactry through one web address, its MCP endpoint:

https://app.artifactry.app/api/v1/mcp

Most apps use OAuth sign-in: you paste the address, a window opens, you sign in to Artifactry and choose Allow. There's no token to copy, and access refreshes itself. Step-by-step guides for each app are in the MCP Integration tab.

What your AI can do once connected

7. Personal access tokens

Some tools can't do a browser sign-in: scripts, scheduled jobs, or an MCP client without OAuth. For those, create a personal access token, a long secret string the tool sends with every request.

1. In the app, open Profile and scroll to Connected AI apps.
2. Choose the app you're connecting (or Other MCP client for scripts) and choose Connect.
3. Tick only the permissions it needs. A read-only script needs just artifact.read.
4. Copy the token straight away. It is shown once and stored only as a hash, so it can't be shown again.

Permissions (scopes)

8. Using Artifactry from the command line

There's no separate CLI to install. The MCP endpoint is ordinary JSON over HTTPS, so any HTTP tool can call it with a personal token. These examples use curl, which comes with macOS, Linux and Windows 10+.

Step 1: store your token

export ARTIFACTRY_TOKEN="paste-your-token-here"

$env:ARTIFACTRY_TOKEN = "paste-your-token-here"

Step 2: list the tools (no token needed)

curl -s https://app.artifactry.app/api/v1/mcp

This returns the tool list and the permission each one needs. It's a quick way to check Artifactry is reachable.

Step 3: search your library

curl -s https://app.artifactry.app/api/v1/mcp \
  -H "Authorization: Bearer $ARTIFACTRY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call",
       "params":{"name":"search","arguments":{"query":"launch plan","limit":5}}}'

List workspaces

curl -s https://app.artifactry.app/api/v1/mcp \
  -H "Authorization: Bearer $ARTIFACTRY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call",
       "params":{"name":"list_workspaces","arguments":{}}}'

Save a page

curl -s https://app.artifactry.app/api/v1/mcp \
  -H "Authorization: Bearer $ARTIFACTRY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call",
       "params":{"name":"save","arguments":{
         "file":"hello.html",
         "title":"Hello from curl",
         "content":"<!DOCTYPE html><html><body><h1>Hello</h1></body></html>"}}}'

Read one artifact

curl -s https://app.artifactry.app/api/v1/mcp \
  -H "Authorization: Bearer $ARTIFACTRY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":4,"method":"tools/call",
       "params":{"name":"get","arguments":{"artifact":"https://app.artifactry.app/a/YOUR_ARTIFACT_ID"}}}'

9. How agents use artifacts across sessions

AI chats forget. Each new conversation starts blank. Artifactry is the long-term memory they can share: one session saves, a later session (even in a different AI app) finds and builds on it.

A typical week

1. Monday, in Claude: "Draft a launch plan and save it to Artifactry." Claude calls save and gets back a permanent link.
2. Wednesday, in ChatGPT: "Find the launch plan in Artifactry and add a pricing section." ChatGPT calls search, reads it with get, then saves the update with create_version. Same link, now version 2.
3. Friday, in Cursor: "Pull the launch plan from Artifactry and turn the checklist into tickets." The coding agent reads the latest version.
4. Any time: "Share the launch plan with Board prep for a week." One share call, inside your organization, no prompt needed.

Every artifact records which AI created or changed it, so "which AI did what, on whose behalf" always has an answer.

10. Troubleshooting

11. Best practices

• Grant the least access that works. A research assistant may only need artifact.read. Add artifact.share only for apps you trust to share.
• One workspace per project or client. It keeps search results focused and makes access easy to reason about.
• Use teams, not email lists. Sharing with "Board prep" stays correct as people join and leave.
• Set expiries on outside shares. A link that ends on its own can't be forgotten.
• Version, don't duplicate. Updating an artifact keeps its history and its link.
• Curate tags. Accept good suggestions and reject the noise; confirmed tags make search sharper for everyone.
• Review connected apps now and then. Disconnect anything you no longer use from Profile.
• Give titles a human could search. "Q3 board deck v2 (final)" beats "deck-final-final.pptx".

12. Limits and file types

• Types: PDF, DOCX, PPTX, HTML, PNG, JPG, JPEG, BMP and GIF.
• Size: up to 25 MB per file.
• HTML preview: Safe Preview (the default) renders HTML and CSS but runs no JavaScript. Owners can switch on Interactive Preview per artifact, which runs scripts in an isolated sandbox with network access blocked.
• Scanned PDFs have no text layer, so they're findable by title, tags and dates rather than their words.
• Thumbnails aren't generated yet; files show a type badge.
• Beta: features may change while Artifactry is in beta. Have an idea? Request a feature.

Before you start

MCP (Model Context Protocol) is a common language that lets AI apps use outside tools. Connecting Artifactry over MCP gives your AI the Find, Save and Share tools from the User Guide.

Every app uses the same address:

You'll need

• An Artifactry account. Create one free if you haven't.
• The AI app you want to connect, signed in.
• About two minutes.

ChatGPT

Prerequisites

• ChatGPT on the web (chatgpt.com) or the desktop app.
• A Plus, Pro, Business, Enterprise or Edu plan. Custom connectors need Developer mode, which isn't available on the free plan. On Business and Enterprise, a workspace admin may need to allow it.

Setup

1. Open Settings → Security and login (on some plans Settings → Apps & Connectors → Advanced) and turn on Developer mode.
2. Go to Settings → Apps & Connectors and choose Create.
3. Fill in the form:
   Connector detailsCopy

   Name:            Artifactry
   Description:     My Artifactry library: search, save and share artifacts
   MCP Server URL:  https://app.artifactry.app/api/v1/mcp
   Authentication:  OAuth

4. Tick that you trust the app and choose Create.
5. A window opens on app.artifactry.app. Sign in with your email code and choose Allow.
6. Start a new chat. Open the + menu, choose Developer mode and switch on Artifactry.

Authentication

OAuth handles everything; you never copy a token. If your company blocks OAuth connectors, ask your ChatGPT admin to approve Artifactry.

Try it

List my Artifactry workspaces.

Write a one-page summary of this chat as HTML and save it to Artifactry.

Find my launch plan in Artifactry and add a risks section as a new version.

Behind the scenes, ChatGPT calls a tool like this:

save({
  "file": "chat-summary.html",
  "title": "Chat summary",
  "content": "<!DOCTYPE html><html>…</html>"
})

Common pitfalls

• No "Create" button: Developer mode is off, or your plan or admin doesn't allow custom connectors.
• Artifactry isn't used in chat: switch it on for that chat from the + → Developer mode menu, or name it: "use Artifactry to…".
• Custom GPTs: GPT "Actions" are a different system and can't use MCP connectors. Use a normal chat with Developer mode.
• Files you attach: ChatGPT can hand attached files straight to Artifactry. Say "save the file I attached to Artifactry".

Claude (Desktop, web and mobile)

Prerequisites

• A Claude account at claude.ai. Custom connectors work on every plan; the Free plan allows a limited number.
• On Team and Enterprise plans, an owner adds the connector once for the organization, then each person connects their own account.

Setup

1. In Claude Desktop or at claude.ai, open Settings → Connectors.
2. Choose Add custom connector.
3. Enter the details:
   Connector detailsCopy

   Name:               Artifactry
   Remote MCP server URL:  https://app.artifactry.app/api/v1/mcp

   Leave the advanced OAuth Client ID and Secret fields empty. Artifactry registers Claude automatically.
4. Choose Add, then Connect. Sign in to Artifactry in the window that opens and choose Allow.
5. In a chat, open the tools menu (the sliders icon) and make sure Artifactry is switched on.

Connectors added at claude.ai also appear in Claude Desktop and the mobile apps.

Prefer a config file? (Claude Desktop)

If you can't use custom connectors, Claude Desktop can reach Artifactry through the small open-source mcp-remote bridge. You need Node.js 18 or newer installed.

1. In Claude Desktop, open Settings → Developer → Edit Config. This opens claude_desktop_config.json.
2. Paste this in (merge with anything already there) and save:
   claude_desktop_config.jsonCopy

   {
     "mcpServers": {
       "artifactry": {
         "command": "npx",
         "args": ["-y", "mcp-remote", "https://app.artifactry.app/api/v1/mcp"]
       }
     }
   }

3. Quit Claude Desktop completely and reopen it. A browser window asks you to sign in to Artifactry and Allow.

Try it

Which Artifactry workspaces do I have?

Turn this conversation into a polished HTML brief and save it to Artifactry.

Search Artifactry for the Q3 finance deck and share it with the Board prep team for 7 days.

share({
  "artifact": "https://app.artifactry.app/a/01J9Y…",
  "team": "Board prep",
  "permission": "view",
  "expires_in_days": 7
})

Common pitfalls

• "Couldn't connect": check the URL ends in /api/v1/mcp with no trailing spaces.
• Claude asks before every call: that's Claude's per-tool approval. Lookups are marked read-only; you can choose "Always allow" for tools you trust.
• New tools don't show up: open Settings → Connectors, disconnect and reconnect Artifactry, then start a new chat.
• Config file ignored: the JSON must be valid (watch for missing commas) and Claude Desktop must be fully quit, not just closed.

Claude Code

Prerequisites

• Claude Code installed and signed in (run claude in a terminal to check).

Setup with OAuth (recommended)

1. Add the server. --scope user makes it available in every project:
   TerminalCopy

   claude mcp add --transport http --scope user artifactry https://app.artifactry.app/api/v1/mcp

2. Start Claude Code and type /mcp. Choose artifactry → Authenticate, sign in in the browser and choose Allow.

Setup with a personal token

For CI or machines without a browser, create a token and pass it as a header:

claude mcp add --transport http artifactry https://app.artifactry.app/api/v1/mcp \
  --header "Authorization: Bearer $ARTIFACTRY_TOKEN"

Or share it with your team in a project .mcp.json (the token stays in each person's environment, not the file):

{
  "mcpServers": {
    "artifactry": {
      "type": "http",
      "url": "https://app.artifactry.app/api/v1/mcp",
      "headers": { "Authorization": "Bearer ${ARTIFACTRY_TOKEN}" }
    }
  }
}

Try it

Save docs/architecture.html to Artifactry in the Engineering workspace.

Find the latest API design doc in Artifactry and check this branch against it.

Common pitfalls

• Shows "needs authentication" in /mcp: choose Authenticate and finish the browser sign-in.
• Token header not working: make sure ARTIFACTRY_TOKEN is exported in the shell that starts claude.
• Check status any time: claude mcp list.

Gemini

Gemini CLI: prerequisites

• Node.js 20 or newer.
• The Gemini CLI: npm install -g @google/gemini-cli, then run gemini once to sign in with Google.

Gemini CLI: setup

1. Add Artifactry with one command:
   TerminalCopy

   gemini mcp add --transport http --scope user artifactry https://app.artifactry.app/api/v1/mcp

   Or edit ~/.gemini/settings.json yourself:
   ~/.gemini/settings.jsonCopy

   {
     "mcpServers": {
       "artifactry": {
         "httpUrl": "https://app.artifactry.app/api/v1/mcp"
       }
     }
   }

2. Start gemini and run:
   Inside Gemini CLICopy

   /mcp auth artifactry

   Sign in to Artifactry in the browser and choose Allow.
3. Run /mcp to see Artifactry listed with its tools.

Gemini CLI with a personal token

{
  "mcpServers": {
    "artifactry": {
      "httpUrl": "https://app.artifactry.app/api/v1/mcp",
      "headers": { "Authorization": "Bearer $ARTIFACTRY_TOKEN" }
    }
  }
}

Gemini Enterprise

Gemini Enterprise asks for an OAuth client ID and secret rather than registering itself. Artifactry can issue one for your organization; during beta, email admin@artifactry.app and we'll set it up. Then add an MCP server in Gemini Enterprise with:

Server URL:         https://app.artifactry.app/api/v1/mcp
Authentication:     OAuth 2.0
Authorization URL:  https://app.artifactry.app/oauth/authorize
Token URL:          https://app.artifactry.app/oauth/token
Client ID / Secret: (the values we send you)

Try it

Use Artifactry to list my workspaces.

Save README.md to Artifactry as an HTML page called "Project overview".

search({ "query": "project overview", "type": "html", "limit": 5 })

Common pitfalls

• Server shows as disconnected: run /mcp auth artifactry again.
• Used url instead of httpUrl: in settings.json, url means the older SSE transport. Artifactry needs httpUrl.
• Tools are asked about every time: approve with "always allow" in the prompt, or add "trust": true to the server entry once you're comfortable.

Cursor, Codex and other MCP clients

Any client that supports remote MCP servers over HTTP works. If it supports OAuth, just give it the URL. If not, use a personal token in an Authorization: Bearer header.

Cursor

{
  "mcpServers": {
    "artifactry": {
      "url": "https://app.artifactry.app/api/v1/mcp"
    }
  }
}

Open Cursor Settings → MCP and choose Login next to Artifactry.

OpenAI Codex CLI

codex mcp add artifactry --url https://app.artifactry.app/api/v1/mcp
codex mcp login artifactry

Check it works

1. Is Artifactry reachable? Open https://app.artifactry.app/api/v1/mcp in a browser. You should see a JSON list of tools.
2. Is your AI connected? Ask it: "List my Artifactry workspaces." You should get your workspace names back.
3. Can it save? Ask: "Save a short test note to Artifactry called Connection test." It should reply with a link starting https://app.artifactry.app/a/.
4. Does it show up? Open app.artifactry.app. The note is at the top of your Library, and Profile → Connected AI apps shows the app as connected.