Connect your agent to Claude

What you get

Your agent already talks to customers all day. This connects that pile of real conversations to your own Claude, so you can ask about it the way you would ask a colleague — and Claude can go and fix what it finds, without you opening a settings page.

It works over MCP (Model Context Protocol), the standard way an AI assistant plugs into an outside system. One connection carries your whole business:

• Agents — your conversations, contacts, channels and the agent's own knowledge.
• Sales — groups, journeys and audiences.
• Helpdesk — tickets, who is waiting and what for.
• Marketing — posts, channels and the strategy behind them.
• CRM — your companies, the people in them and the fields you defined.
• Booking — services, opening hours and the diary.
• The task board — anything waiting on a decision.

Install the plugin

One install, and Claude has your whole business: every app on one connection, plus the operator that knows what to do with them. Two minutes, and there is nothing to set up on our side first — no account, no agent, no organization. The connect screen builds the workspace as you go.

Claude’s own menus are in English even if you use AssistantLabs in Hebrew, so the button names below are written the way they appear on screen.

Same screens on claude.ai and in the Claude desktop app — they are the same settings. Install once and it follows your account.

We are in Claude’s own plugin directory, so there is no address to paste and nothing to download. Open Customize in the left sidebar, choose Plugins, then Discover, and search for:

Click Add. The plugin moves to Yours, and it is on your account in every Claude you sign into.

Working in a Claude Code terminal? Install from our repository instead, with these two commands.

Sign in — the step that is not where you expect it

You installed a plugin and were never asked to log in. Most people read that as a failure and start over. It is not: the sign-in comes later, and you are the one who triggers it.

Start a new chat and type this. It works out where you are — brand new, half configured, or running for months — and asks for whatever it is missing.

The first thing it needs is your account, so it hands you three steps and a link. Follow them in another tab and come back.

• Open claude.ai/customize/connectors.
• Click +, then Add custom connector.
• Paste the address below and click Add. Leave the OAuth fields empty.

Why the plugin cannot do this for you: on claude.ai a plugin runs in a sandbox that cannot open a browser on your machine. In a Claude Code terminal it can, and /al signs you in there without any of this.

Clicking Add brings up our connect screen. This is the whole decision, and it is the only one you make:

• No account yet is fine. If you have never used AssistantLabs, this screen asks what your business is called and creates the workspace. There is nothing to set up first.
• No agent yet is fine either. The board, the CRM and the diary belong to the business, not to an agent — you can connect now and build the agent with Claude afterwards.
• One connection, every app. Agents, Booking, Sales, Helpdesk, CRM, Marketing and the board are all on this one address. What Claude can reach is what you tick here, and you can change it without connecting anything again.
• Your account, your permissions. No key, no password, nothing pasted anywhere. The permissions that reach a real person — messaging a customer, taking a journey live, publishing a post — are marked, and never ticked for you.

What to ask it

The connection is not a chat window onto your database — it is a set of tools Claude picks from. You describe what you want in ordinary language and it works out which ones to call.

Reading what actually happened

• “What did customers ask me this week?”
• “Which conversations got no reply?”
• “What time of day do most messages come in? Should I be staffing differently?”
• “Read the last 50 conversations and give me the five questions the agent answered badly.”
• “Has anyone complained about delivery times this month?”

Fixing the agent while you talk to it

This is the part people underestimate. Claude does not just report — with the edit permission on, it can change what the agent knows and the change applies to real conversations from that moment.

• “Customers keep asking whether we ship to Eilat. Add it to the FAQ and show me exactly what you changed.”
• “The agent is too formal. Loosen the tone but keep it polite.”
• “Re-read our website and update the business details from it.”
• “Simulate a customer asking about returns and show me what the agent says.”
• “File a training suggestion for that bad answer instead of changing it live — I want to review it first.”

Training suggestions are review-first: they land in the Training tab for you to approve. Nothing goes live on its own.

Contacts, channels and templates

• “How many new contacts came in this month, and where from?”
• “Which channels are connected right now?”
• “Draft a WhatsApp template apologising for a delayed order, then create it for me to submit.”

On very large questions: answers are capped so a single question cannot pull an unbounded amount of transcript — roughly 300 conversations per analysis and 200 messages per conversation. When a result is cut short it says so and asks you to narrow the dates, rather than quietly answering from half the data.

Sales, CRM and the board

There is nothing more to add. One address carries all of it — what you see is decided by what you ticked when you signed in, and you can change that any time without connecting anything again.

Sales

Groups, journeys and audiences: “which journeys are running and how are they doing?”, “build me a win-back journey for customers who have not ordered in 90 days”, “who is in the hot segment right now?”. Planning and drafting are granted by default; taking a journey live is not — that sends real marketing messages and cannot be recalled, so it is a separate tick.

CRM

Your companies and the people in them: “which accounts have no phone number?”, “who did we speak to at the plumbing firm?”, “fill in what is missing on this company from their website”. Reading is granted by default, and so is filling in blanks — enrichment never overwrites something a person typed unless you ask for it out loud. Changing how a page is laid out is a separate tick: a page is the app for everyone in the workspace.

The task board

“What is stuck?”, “open a task to call the supplier back, due Thursday”, “summarise everything assigned to me this week”. Both reading and writing are granted by default here — nothing on a task board reaches a customer.

Did not tick one at the time? Say so in the chat and it will point you back at the connect screen. The products you left off are one tick away, not another connector.

Disconnecting, and when it breaks

Access you cannot take back is not access you should have given. Both paths have a one-click undo, in the same panel you connected from.

• Connected apps on the connect page lists who is connected, what they may do, and when they last used it. Disconnect takes effect on that app's very next request.

• The connect screen says you have no agents — that is a note, not an error. The board, the CRM and the diary belong to the business, so approve anyway and build the agent with Claude afterwards.
• Claude says it is not authorised — the key was revoked, or the sign-in was disconnected. Reconnect; nothing is lost.
• Claude refuses one particular request — that permission was not ticked at approval time. Disconnect and reconnect, ticking it.
• The connector was added but the chat ignores it — it is off in that conversation. Click the plus next to the message box, choose Connectors, and switch it on. It is per-chat, not global.
• You pasted the address and nothing happened — check you copied the whole thing, including /mcp at the end.

Next, the same job on the other side of the fence: connecting to ChatGPT and Codex.