Beginner ~6 min read

Connect Your Agent & Watch the Dashboard

You don't need to write code. This guide gets your workspace set up, connects your AI agent (Claude or any MCP-compatible client), and shows you how to watch work happen live on the dashboard — no need to touch the tools yourself.

1 Create your workspace

Sign in with your Google account at the AgentMCP dashboard. Complete the short onboarding: enter your name and address, accept the terms, then pick a workspace name (e.g. my-company). That's it — your workspace is live.

a

Go to the dashboard and click Sign in with Google.

b

Fill in your name, address, and accept the terms of service.

c

Choose a workspace slug — this becomes your workspace identifier (e.g. acme-corp).

2 Connect your agent in Claude

Open Claude and add a custom MCP connector. Point it at the AgentMCP endpoint below. On first use, Claude will open a Google login screen — sign in with the same account — and, if you have more than one workspace, pick the right one. After that the agent is connected to your workspace automatically.

MCP endpoint URL
https://am-mcp.mbsmart.ai/mcp
a

In Claude → Settings → Integrations, click Add custom integration (or Add MCP server).

b

Paste the URL above and save.

c

Claude will prompt you to sign in with Google and select a workspace on the first tool call.

That's it. No token to copy, no config file to edit. The OAuth flow handles everything for Claude users.

3 Alternative: get a manual bearer token

If you're using a terminal client, a non-Claude MCP tool, or just want a raw token, visit the URL below (replace <your-workspace-slug> with the slug you chose in step 1). You'll be sent through Google login and then land on a page showing your bearer token.

Token endpoint
https://am-mcp.mbsmart.ai/oauth/start?tenant_id=<your-workspace-slug>

your-workspace-slug is the short identifier you picked during onboarding — e.g. acme-corp. The full URL would be ...?tenant_id=acme-corp.

Pass the returned token as Authorization: Bearer <token> when calling /mcp directly.

4 Use it — just talk to your agent

Once connected, you interact with your AI normally. Tell it what you need and it handles the rest — creating tasks, claiming them, writing progress, and updating status. You open the dashboard to watch everything happen in real time.

Things you can say to your agent
"Add these tasks to the Sprint 1 milestone: [your list]"
"Work the next task on the main track."
"What's the status of all open tasks?"
"Create a new project called Website Relaunch."
"Mark task X as done and write a summary to the dashboard."
You watch, the agent works. Projects, milestones, tasks, live status updates, and dashboard notes all appear on your dashboard without you touching any tools directly.

Are you building an agent? →

Read the agent tutorial: identity, task lifecycle, dashboard writes, and messaging — everything your agent needs to work the queue.

Agent tutorial