Bring a model provider
Use an available provider, or connect one already in use. Configure models around the agents’ work.
Explore agent setupWork with agents connected to business information and tools. Develop applications, schedule recurring tasks, and share useful capabilities across the team.
Review the customer feedback and flag the issues that need attention.
Explore themes and inspect original comments.
Describe what the team needs help with and how the work should be done. Configure the agent through conversation, or adjust its instructions, skills, and model in settings.
Define the requirements and checks that should guide the agent.
Choose from available skills. Selected skills stay attached to the agent.
Use an available provider, or connect one the team already uses.
Skills can be installed and updated from a Git-based catalog. Provider connection details and model settings belong in the configuration guide.
Skills
Model
OpenRouterSuggested
Start with the integration catalog, or connect internal systems and services. Give each agent the information and actions relevant to its work.
Connect business services from the catalog.
Bring internal services through MCP, APIs, and custom tools.
Choose which capabilities are available to the agent.
MCP connects agents to compatible tool servers. APIs and custom tools can support additional services. Authentication and setup requirements vary by connection.
Illustrative catalog marks · connection methods depend on configuration
Agents can use tools, delegate to specialists, and work in a browser or computer environment. Follow the activity and review the resulting work.
Prepare a brief using the latest account notes.
Recent changes, open issues, and questions to confirm.
See which tools the agent uses and what they return. Review the work with its supporting information.
The support evidence and account context come together here.
A coordinating agent can delegate parts of a task to other agents and bring their contributions together. Independent work can run in parallel.
Return to the conversation and supporting evidence.
Agents can interact with web pages and computer tools. Observe the work and return to the conversation it supports.
Browser and computer tools operate within configured execution environments. Sandbox lifecycle, persistence, and isolation details should be checked for the deployment. Enterprise documentation covers these choices.
Agents can develop custom tools and lightweight applications to support the work, with an interface and the logic behind it. Describe what is needed and refine it through conversation.
Reusable ways to retrieve information, perform actions, and follow a process.
Controls and interfaces for exploring information and carrying out tasks.
Adjust the application when requirements change.
Make this an application. Let the team explore themes and open the original comments.
The email names a different screen.
Turn a request into a one-time or recurring task. For more involved work, combine tool actions and agents in a multi-step pipeline.
Specify when the work runs and which agent handles it.
Coordinate tools, agents, and conditional steps.
Inspect status, timing, and the conversation produced by each run.
Collect records, prepare context, ask an agent to review, then save the result. Pipelines can include tool steps, other agents, and conditional logic. Scheduling and external actions require the relevant configuration.
Run this review every Monday morning.
Use an available provider, or connect one already in use. Configure models around the agents’ work.
Explore agent setupStart with the catalog, or connect internal systems, services, and specialist tools.
Explore connectionsExplore deployment, isolation, access, and support options for the organization.
Enterprise optionsBring a task or process the team already handles. We will walk through the agents, connections, applications, and recurring tasks that could support it.