Last updated:
AI
Configure how Agent0 responds in your Dash0 organization. Access these settings at Settings → Agent0.
Agent0 Settings
Default Model
Sets the model level and reasoning effort Agent0 runs with when nothing more specific applies. The table has one row per Agent0 surface: Agent0 Chat, Automations, Slack, and MCP.
Leaving a row on Default inherits your organization's default for that surface, or the Dash0 default if none is configured. It does not pin the value it currently resolves to, so later changes to the inherited default reach you. Every row can still be overridden within a thread or an automation. Slack answers link here from the change model link in their footer.
Higher levels and higher efforts consume more credits per turn. For what each level is for, what it costs, and how a level and an effort resolve against each other, see Model Selection.
Infer Context
When enabled, Agent0 automatically infers context from the specific page you're looking at when opening the Agent0 search, and passes it to Agent0 to provide you with contextualized results. (Recommended)
This setting allows Agent0 to automatically gather relevant traces, metrics, and logs based on your current view, making its analysis more accurate and relevant to your investigation.
Write Action Approval
Controls whether Agent0 asks for your approval in Chat before it runs an action that creates, changes, or deletes something.
- Off (default): Agent0 runs write actions as part of answering your request.
- On: Agent0 pauses for covered write actions unless you already approved that tool for the thread. It shows what it wants to do and waits for approval or denial in the conversation.
The setting is admin-only, and it is disabled while Agent0 itself is off. GitHub actions are not covered by it.
For the approval card and what approving or denying does, see Approve Write Actions.
Agent0 via MCP
Controls whether Agent0 can be invoked through the Dash0 MCP Server, so an AI assistant such as Claude Code or Cursor can hand it a whole investigation instead of querying telemetry itself.
- On (default): assistants connected to the Dash0 MCP Server can delegate to Agent0. Each delegation consumes AI credits.
- Off: the delegation tools (
runTask,waitForTask, and the thread inspection tools) are removed from the MCP Server for everyone in the organization. The read-only telemetry tools stay available and remain free.
This is the organization-wide way to guarantee that MCP usage does not consume AI credits, and it applies no matter which client each person connects with. The setting is admin-only, and it is disabled while Agent0 itself is off.
For what delegation does, what it costs, and how an individual can disable it in just their own client, see Delegate to Agent0 via MCP.
This setting is about the Dash0 MCP Server, which exposes Dash0 to your AI assistant. It is unrelated to Settings → Integrations → MCP, which connects external MCP servers into Agent0 as connectors.
Setup Script
A Bash script that runs once in each new Agent0 chat sandbox, before Agent0 starts, to pre-install CLIs, runtimes, or dependencies. There are two:
- Your preferences → Setup script is your own. It runs in your chat threads and replaces the organization's script for you.
- Organization settings → Setup script runs for every member who has no script of their own. It is admin-only.
When an admin turns on Make setup script mandatory, the organization's script runs first for everyone, followed by the member's own script in the same shell. Setup scripts apply to Agent0 Chat only, not to automations, Slack, or Agent0 via MCP, and a change reaches threads started after it is saved.
For precedence, network access, writing a reliable script, and seeing what ran, see Use a Setup Script in Chat.
Auto-share Memories
Controls whether Agent0 can automatically share the memories it saves while working for you in Dash0, including the side panel. Sharing is limited to the memory's organization and dataset. Find the setting under Your preferences → Auto-share memories.
- On (default): Agent0 can share a memory when it judges the memory useful to your organization and free of anything private. Memories saved in Slack, Linear, or over MCP require you to share them.
- Off: Agent0 does not automatically share new memories. Memories you already shared stay shared. You can still share or stop sharing a memory yourself from the Memories page.
The setting is personal: it applies to your memories only, and every member sets their own.
For who can use a shared memory and how to share or stop sharing one, see Control Memory Sharing and Access.
Further Reading
- Agent0 Overview — AI-powered observability assistant
- Model Selection — Model levels, reasoning effort, and what each costs
- Chat with Agent0 — Interactive investigation interface
- Agent0 Automations — Automated issue detection and response
- About Memories — What Agent0 remembers across sessions, and who it is shared with
- Approve Write Actions — Requiring approval before Agent0 changes anything
- Delegate to Agent0 via MCP — Handing investigations to Agent0 from an AI assistant, and what it costs
- Use a Setup Script in Chat — Pre-installing tools for Agent0 Chat, and which script runs when


