> ## Documentation Index
> Fetch the complete documentation index at: https://java.agentscope.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents: reusable capabilities

[简体中文](/v2/zh/service/agents)

**DESIGN → Agents** manages identity, behavior and execution. Reuse an Agent in Chat, Issues, Teams and Endpoints. Saving its definition does not start work.

## Choose an execution type

| Type     | Where it runs                                     | Onboarding                                               |
| -------- | ------------------------------------------------- | -------------------------------------------------------- |
| Managed  | Service Harness and Dataplane                     | Create an Agent using AgentScope Managed                 |
| Hosted   | A Coding Agent provider on your machine or server | Connect a Runtime Host, then select a discovered Runtime |
| External | Your independently deployed application           | Register through the SDK to enter the Agent catalog      |

The work entry points are shared, but models, tools, recovery and configuration projection differ. Catalog visibility does not guarantee every conversation or dispatch capability. See the dedicated [Managed](/v2/en/service/managed-agent), [Hosted](/v2/en/service/hosted-agent) and [External](/v2/en/service/external-agent) guides.

## Create an Agent

1. Start creation and enter its name, purpose and Instructions.
2. Link a Workspace when it should reuse operating guidance, skills, tools or subagent definitions.
3. Explicitly choose Runtime under Execution. An online Hosted provider may be selected initially; choose **AgentScope Managed** for a Managed Agent.
4. Leave Model empty for the runtime default, or enter a model identifier supported by that provider.
5. For Managed Agents, choose an Environment in Advanced settings. The automatic local default requires administrator permission for Local execution.
6. Select **Create & open agent**, then review the resulting configuration.

Agent key is a stable identity within the scope. The display name communicates purpose to colleagues.

## Configure the detail pages

| Page                   | Purpose                                  | First verification                             |
| ---------------------- | ---------------------------------------- | ---------------------------------------------- |
| Behavior               | Responsibilities, instructions and model | Ask a bounded question                         |
| Workspace              | Materials and execution resources        | Read a known file                              |
| Skills                 | Reusable procedures and supporting files | Run a task matching a skill                    |
| Tools                  | Tools and MCP connections                | Make a read-only call and check authentication |
| Subagents              | Delegated specialist capabilities        | Inspect a delegated result                     |
| Versions               | Definition history                       | Use new work to verify the intended version    |
| Connections → Channels | Messaging bindings                       | Send a request through the actual channel      |

## Add capabilities incrementally

Verify conversation first, then file access, an external tool and a specialist skill. Store credentials in [Vault](/v2/en/service/vault) and shared knowledge in [Memory](/v2/en/service/memory). Configuration does not automatically install executable tools or grant external permissions.

A shared Workspace can affect multiple Agents. Check its consumers before editing and verify with a new Chat or Issue. Do not assume that running work switches definitions immediately.

Next: [Teams](/v2/en/service/teams) · [Workspaces](/v2/en/service/workspaces).