Artala's connector lets Claude, ChatGPT and other AI assistants read your workspace and make changes in it: what's overdue, who has room next week, how the sprint is going — then create the task, tick the checklist, log the time. In Arabic, Hebrew, Urdu or English, with right-to-left text kept intact.
Ask in your own words and your own language. Your assistant looks up the real names of your boards, people and statuses before it answers.
What's overdue on the Marketing board, and who owns it?
Who has capacity next week for another three points of work?
How is the current sprint going? Anything stuck in review for more than three days?
Add a task for Dana to review the Q3 budget, due Friday, high priority.
Mark the homepage copy task done and tick off “send to legal” on its checklist.
Log an hour and a half on the logo redesign for Omar, note “client call”.
Run our client-onboarding playbook on the Sales board for Acme.
How do I invite someone as a Viewer?— answered from Artala's own user guide, with the button names as your screen shows them.
Nineteen tools: twelve that read, seven that make changes. Nothing can be deleted through the connector.
Answers come from your live workspace, not a copy.
Only if you allow changes when you connect — and your assistant asks first, as its settings say.
The connector goes through the same checks as the app — it isn't a back door.
Untick Allow changes when you connect and the assistant can only read. With changes allowed, Claude asks you to approve each one by default (on Claude Team and Enterprise, your organisation's owner controls that); in ChatGPT, the connection's Permissions setting decides when it asks.
Each person signs in as themselves: the assistant sees what they see and changes what they could change in the app. A board where they are a Viewer can't be changed through the connector either.
Changes appear in task history, the audit log and notifications under the person's name and the assistant's — for example “Dana Cohen (via Claude)” or “Dana Cohen (via ChatGPT)”. Webhooks and automations fire exactly as they do for a change made in the app.
A change to an existing task needs its id and its current title, so the assistant can't edit the wrong one. The same request sent twice in a row isn't applied twice, and a description rewrite that would drop a pasted image is refused.
Arabic, Hebrew and Urdu names and titles come back marked so they display in the correct order, even inside an English sentence. Dates follow your workspace's time zone, the same as Reports.
Artala runs no AI model for the connector. It answers the tool requests your assistant sends and never receives your conversation. Your usage is covered by your plan with your AI tool.
About two minutes. You need an Artala Pro workspace. Each person signs in with their own Artala account — no key to create or share. The server URL is the same everywhere: https://app.artala.app/api/mcp
Web and desktop
Artala, the server URL above. Leave the authentication settings as they are, and save.Started from the desktop app? At the end your browser asks whether to open Claude — choose Open Claude. If you cancel, just click Connect again.
Set up on chatgpt.com · ChatGPT calls these plugins
Artala, Server URL as above, Authentication OAuth. Leave the advanced OAuth settings as they are.Developer mode depends on your ChatGPT plan and, on a business account, your workspace's settings.
Add it in ChatGPT in your browser. It then appears in ChatGPT's desktop and mobile apps by itself — the desktop app's own MCP settings are for local servers and can't sign in to Artala.
In your terminal
/mcp in Claude Code, choose artala and authenticate.Add --scope user to use it in every project. Other assistants that sign in with OAuth connect the same way: give them the server URL and follow their sign-in.
Before anything connects, Artala shows you which assistant is asking, who published it, the workspace and account you are signed in with, and where you'll be sent back to. Check them, then click Connect.
claude.ai, chatgpt.com) and start with Allow changes ticked.For tools that can't sign in — or to give a whole Claude organisation one connection — use a workspace API key. An Owner or Admin creates one in Settings → API Keys (name it after the tool, for example Claude connector). The connector then sees what the key's creator sees, for everyone who uses it, and changes appear as “Dana Cohen (API: Claude connector)”. ChatGPT accepts sign-in only, not keys.
authorization and enter Bearer tb_… — the word Bearer, a space, then your key. Request headers are a Claude beta available to a limited set of organisations.tb_…) without the word Bearer. On an organisation account, an admin has to allow custom connectors first.Any MCP client that can send an Authorization header can connect with a workspace API key.
Keeps the key out of the file: set ARTALA_KEY in your environment.
VS Code asks for the key once and stores it securely.
Pro. If your workspace moves off Pro, the connector stops working until the plan is restored.
Claude (web and desktop), Claude Code and ChatGPT sign in with your Artala account, and so can other MCP clients that sign in with OAuth. Perplexity, Cursor, VS Code and other clients that can send a header connect with a workspace API key.
When you sign in, yours, with the assistant's — for example “Dana Cohen (via ChatGPT)” — and “my tasks” means your tasks. With a shared API key, it's the key creator's name and the key's — “Dana Cohen (API: Claude connector)” — whoever asked; that connector can't tell who is asking, so the assistant asks for your name when it needs it.
No. The connector has no delete tools. It can clear a field (a due date, an assignee) or untick a checklist item when you ask it to.
Yes. Ask in any language; statuses and priorities can be named in any of Artala's interface languages, and people's names are matched as they are stored, in any script. Right-to-left values come back marked so they display in the correct order inside your answer.
Allow changes was unticked when you approved the connection, so ChatGPT connected read-only — it says that not all requested permissions were granted. Reconnect and tick Allow changes on Artala's approval page. ChatGPT's own Permissions setting only decides when it asks you; it can't grant changes Artala didn't allow.
Disconnect it in Artala under My Profile → Connected AI assistants — it stops at once. “Sign out of all devices” and a password reset disconnect every assistant too. For a shared key, disable or revoke it under Settings → API Keys; the connector stops within about a minute. You can also remove Artala from your assistant's own connector or plugin settings.
Yes. Connector requests count toward your workspace's API rate limit, like any other API request.
Artala is free for up to 3 users. The AI connector comes with Pro.