Core concepts

Follow a project assistant through one conversation. This example uses Service, the shared platform accessed through Console or the HTTP API. For embedded or interactive execution, see Choose your path.

Configure the assistant

Your organization gives you access to a workspace, which contains your team's agents, resources, and conversations. Your roles in the organization and workspace determine what you can use or change.

An agent combines a model, instructions, and tools. For example: “Summarize project progress and cite the information you used.” Each saved configuration is a revision; each run uses a selected revision.

Component	In this example
Model	Interprets requests and produces responses or tool calls
Tool	Looks up a task, reads a file, or performs another configured action
Connection	Gives access to external tools through MCP or an app account
Environment	Provides the files and command execution used by tools
Memory	Holds project decisions the agent can read and update across conversations

Start with a model and instructions. Add access to your project data before asking the assistant to look up real progress. Resource basics explains the setup.

Start a conversation

Ask: “Summarize the project status and list the next two tasks.”

A session groups the conversation's threads. Each thread has its own history and inbox. Forking creates another thread in the same session to explore a different direction. A run advances a thread using the agent's model and tools.

---
config:
  flowchart:
    useMaxWidth: false
  sequence:
    useMaxWidth: false
---
flowchart TB
    Message["Send a message"] --> Run["Agent works on the request"]
    Run --> Calls["Model and tool calls"]
    Calls --> Result["Answer or request for input"]

You see output as it streams. A message sent during execution can guide the active run or wait in the inbox for later work.

Answer a question or approval

If the assistant asks which project you mean, answer in the conversation. If a tool needs approval, approve or reject it. The conversation continues with your response. Applications handle these through Waits, approvals and questions.

Continue later

Ask “Which task should I do first?” in the same thread to continue from its history. Service keeps running when you disconnect and stores the conversation for your return.

Information	Purpose
Conversation history	Continue this discussion
Memory	Save and retrieve information across conversations
Environment files	Store documents and other working files for tools

Attach memory to share saved decisions across conversations. It can supply file context or recall relevant records automatically; enabled memory tools let the agent search and update it. Environment configuration determines how long its files remain available.

Next, try an agent in Console or connect your application. See Agents, threads and runs for revision selection and execution details.