Manage agent settings ↗

Documentation Index#

Fetch the complete documentation index at: https://docs.langchain.com/llms.txt Use this file to discover all available pages before exploring further.

Manage your agents in Agent Builder.

This page explains how to manage the settings for your agents in Agent Builder.

Change the model#

To change the model for your agent:

1. In the LangSmith UI, navigate to your agent’s inbox.
2. Next to the agent name, click the Edit Agent icon.
3. In the top right corner, click the Settings icon.
4. Select the Model you want to use.
5. Enter the API key for the model.

For information on how to add a custom model, see Custom models.

Reconnect tool integrations#

To reconnect a tool integration to an agent:

1. In the LangSmith UI, navigate to your agent’s inbox.
2. Next to the agent name, click the Edit Agent icon.
3. In the top right corner, click the Settings icon.
4. In the Connected Integrations section, click the Connect button next to the tool you want to reconnect.

Download agent files#

To download the files for your agent, click the Settings icon in the top right corner of the agent and select Download ZIP. This will export the agent configuration as a ZIP file.

Change access to the agent#

Agents can either be private to the creator or shared within a LangSmith workspace.

To change the agent visibility, click the Visibility settings icon in the top right corner of the agent and select either Only me or Workspace.

Workspace-scoped agent details#

While workspace-scoped agents are shared, some details are public, while others are private:

• Threads are always user-scoped, so even if an agent is workspace-scoped, the chat history created within that agent will always be private and only accessible to the specific user who created them.
• The system prompt, selected tools, and sub-agents will be public on workspace-scoped agents. Users will not be able to modify these fields on the original workspace-scoped agent, but can make changes once they’ve cloned the agent.
• The trigger type on workspace-scoped agents is public (for example, Slack message received), but the specific connection with the trigger (for example, the Slack channel, or Gmail address) is not shared. This way, users know what trigger to use when cloning an agent, but can’t gain unauthorized access to any connections the original user has set up.

Update memory#

Your agent can remember information from previous conversations and use it to make better decisions in future conversations. By default, your agent will require approval to update its memories folder.

To disable this requirement for your agent:

1. In the LangSmith UI, navigate to your agent’s inbox.
2. Next to the agent name, click the Edit Agent icon.
3. In the top right corner, click the Settings icon.
4. In the Memory section, toggle Require approval to update memories to off.

Use the agent programmatically#

You can use the LangGraph SDK to connect to your agent through code. To view the code snippets needed to call your agent programmatically:

1. In the LangSmith UI, navigate to your agent’s inbox.
2. Next to the agent name, click the Edit Agent icon.
3. In the top right corner, click the Settings icon.
4. Click the View code snippets button.
5. Copy the pre-populated code snippets for your agent.

For more information, see Call agents from code.

Pause agent#

To pause an agent, pause its trigger:

1. In the LangSmith UI, navigate to your agent’s inbox.
2. Next to the agent name, click the Edit Agent icon.
3. In the graph view, click the Pause button in the Triggers box.
4. Click Save Changes.

To resume, click the Resume triggers button in the Triggers box.

Delete agent#

To permanently delete an agent:

1. In the LangSmith UI, navigate to your agent’s inbox.
2. Next to the agent name, click the Edit Agent icon.
3. In the top right corner, click the Settings icon.
4. Click the Delete Agent button.
5. To confirm the deletion, click the Delete button.

This action cannot be undone. It will permanently delete the agent, all threads linked to the agent, and unlink any attached triggers.

Edit this page on GitHub or file an issue.

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.