Skip to content

Concepts

superglue organizes work into a few building blocks. The sidebar follows the same order: agent sessions and projects at the top, then the brain, the capabilities, and the control panel.

Agent sessions

A session is a conversation with the agent. Sessions can belong to a project or run on their own.

You describe the outcome. The agent plans, reads client data, builds and tests tools, and asks for approval before writing to a client system.

Projects

A project is one engagement. It holds tasks with status, owners and due dates, and it links the tools and files involved. Sessions started in a project prioritize that context. See Project setup.

Brain

The brain is what superglue knows about your work. It holds:

  • Connected context sources such as meeting notes, Slack, Teams and email
  • Uploaded files such as client exports and your templates
  • Playbooks that hold your methods
  • Artifacts the agent has produced

See Project context.

Connections

A connection is one authenticated account for an API, database, file server or other service, for example “Contoso sandbox”. It stores the URL, the authentication method, the secret values, an optional secure gateway and usage notes.

Secret values are encrypted and never appear in agent prompts. To connect a service, describe it to the agent. The agent configures the connection and asks for secret values in a secure form. Manage connections and the sign-in apps they reuse on the Connections page.

Tools refer to the system, such as NetSuite, not to one connection. Technical pages call this the system ID, for example netsuite. The same tool runs against a sandbox or a production account without changes. A run uses the connection selected for it, or your default connection for that system.

Tools

Tools are saved, repeatable actions: a migration load, a nightly sync, a reconciliation query. The agent builds and tests them with you.

A saved tool runs the same steps every time with no AI at execution time. Tools run on demand, on a schedule, from a webhook, from the API, or through an MCP server.

Artifacts

Artifacts are the documents the agent produces: data assessment reports, mapping proposals, questionnaires, reconciliations and spreadsheets. They live in Brain next to your uploaded files. Documents download as PDF, Word or Excel, spreadsheets as Excel, and decks as PDF or PowerPoint. Artifacts can be shared with a client through a link.

Runs

Every tool execution creates a run with its trigger, user, input, timing and errors. With Store run results on in Settings, a run also keeps its step results and output. Runs are the audit trail and the starting point for investigating a failure. See Monitoring runs.