Make your first connection
Go from a connected system to an answer grounded in your infrastructure.
You need a ConvergeSense workspace, an owner who can configure connections, and an Engine that can reach your systems. If your team already has an Engine and approved tools, skip to Connect your agent.
Prepare your workspace
Sign in to ConvergeSense. Use the workspace that owns the systems you want to inspect. Your account role controls what you can configure and what your agent can access.
Owners manage Engines, connections, and tool policies. An agent connection never grants more access than your account and the workspace policy allow.
Enroll an Engine
Open Connector Engines in the platform and use the enrollment flow. Install the Engine on a host that can reach the intended systems, then follow the generated enrollment instructions.
Confirm that the Engine is online and sending a recent heartbeat before adding a connection. Device credentials and raw sessions stay on this host. See the Engine guide for the connectivity boundary.
Add a connection
Open Connectors, then Add connection. Select your Engine and a reviewed connector release. Follow the connection wizard to test the target system and activate the connection after its checks pass.
An owner then approves the intended read tools and enables them for the appropriate role and client. A connected device alone does not make every operation available to an agent.
Connect your agent
Add this remote MCP endpoint to your client:
https://mcp.convergesense.com/mcp
Complete the browser sign-in, select one workspace, review the permissions and data-sharing notice, and approve the connection. See agent setup for supported clients.
Ask a first question
Start with a question that lets the agent inspect its available context:
Which ConvergeSense workspace am I connected to,
and which approved tools can I use?
Then ask a focused question about a system or exact identifier you are authorized to inspect. The agent discovers an approved tool, reads its input contract, runs it, and polls if the result is pending.
Check the answer's evidence, observation time, freshness, and warnings. A successful read means the read completed; it does not by itself mean the system is healthy.
Nothing showing up? An online Engine, a ready connection, an approved tool, and a valid agent grant are separate requirements. Start with the troubleshooting guide.