SigNoz Cloud dashboards are now agent-native

TL;DR

Agents are creating a growing share of SigNoz Cloud dashboards as observability becomes agent-driven.

A few months ago, we shipped agent-native observability with the SigNoz Cloud MCP server and Noz, our in-app AI assistant. As part of building observability for the AI era, we’re going a level deeper by redesigning the dashboard data model with agents as first-class users.

The SigNoz Cloud dashboard schema is now structured and strictly validated. Agents write against a defined schema instead of inferring field names. Creating and editing dashboards is faster, more reliable, and lighter on tokens.

We’ve also revamped the dashboard list with custom views, pinning, and search powered by a query DSL, making large dashboard collections easier to organize and navigate.

Let’s look at both changes.

An agent-native dashboard schema

As more agents started operating dashboards, we found the existing JSON was not fully optimized for agents. It was largely free-form and worked well with the UI as the consumer. For an agent, inferring the schema, constructing or modifying JSON, and validating a change could become a bit challenging.

The new SigNoz Cloud dashboard schema treats agents as first-class users. Agentic operations on dashboards are now faster, more reliable, and lighter on tokens.

1. A structured, validated schema. The schema now has a defined hierarchy, with a place for every field. An agent knows where to look instead of inferring. Creating a panel, editing a query, or adding a variable means writing to a known path, validated on the way in.

dashboard
├── tags              key:value pairs on the dashboard
├── sections          groups of panels
│   └── panels        one visualization each
│       └── queries   what data the panel fetches
└── variables         filter across panels

The new dashboard schema

2. An open format. The new schema is built on the open CNCF Perses dashboard specification, extended with SigNoz-native plugins for queries and panels. Existing Perses dashboards can be migrated with a lightweight conversion, and your dashboards aren't locked into a proprietary format.
3. Dashboards as code. A built-in editor lets you read, edit, and format a dashboard's JSON in the app, with the panel updating as you change it. Agents can edit dashboards using the same JSON, and can change many panels in a single request. Panel styling is part of the schema too, so a fill mode change is one field.

4. More efficient edits by agents. An agent now sends only the field it is changing, whether that is updating a filter inside a panel query or adding a new panel. That means a single PATCH operation instead of rewriting the entire dashboard. Smaller writes use fewer tokens and reduce the risk of agent errors.

curl --location --request PATCH '<YOUR-SIGNOZ-URL>/api/v2/dashboards/<ID>' \
--header 'SIGNOZ-API-KEY: <YOUR-API-KEY>' \
--data '[
    {
        "op": "replace",
        "path": "/spec/panels/196021a6-3b94-454c-9c9d-2b4170708038/spec/queries/0/spec/plugin/spec/queries/0/spec/filter/expression",
        "value": "service.name = '\''signoz'\''"
    }
]'

One PATCH request updating a single query filter

5. JSON and UI always in sync. Edit in the builder and the JSON updates. Edit the JSON and the dashboard re-renders. There's no import step, so an agent always works against the current state.
6. Terraform with better drift detection. signoz_dashboard is now a typed spec tree instead of opaque jsonencode(...) strings. The provider validates structure at plan time rather than accepting arbitrary JSON, which cuts configuration drift. A script handles the state migration from the old resource.

If you're moving to SigNoz Cloud from another platform, you can use agents to migrate your dashboard estate.

We've also updated every dashboard template in our dashboard template in our library to the new schema, so you can start from a template and use an agent to edit it.

Managing hundreds of dashboards is now easier

As your product scales, the number of dashboards can increase, and you may end up with hundreds of them. At that scale, finding and organizing the dashboard you need can get painful.

So alongside making dashboards agent-native, we revamped the dashboard list. Organizing and finding a dashboard across hundreds of them is now easier.

1. Group dashboards into custom views, like team views. Some teams want to see only their own services' dashboards. A single team may want several views, one for services and one for databases. Until now, everyone got the same list. Now, you can create custom views from queries over name, creator, or any tag key you use. Views are shared across your organization.

2. A new search, using query DSL. Finding a dashboard used to mean remembering its name, which gets harder as the list grows. Now, search dashboards by writing a query, narrow the list however you need. For example, team contains 'alpha' narrows to one team's dashboards. A search result can be saved as a custom view in a single click.

The same DSL backs the list API, with limit, offset, sort, and order, so a script or an agent can narrow to the dashboard it needs instead of listing everything.

3. Pin the dashboards you open daily. Pinned dashboards move to the top of your list. Pins are personal, so every user keeps their own set, and they apply across all your views.

4. Clone dashboards, panels, and sections. If you want a dashboard close to one you already have, clone it and make the changes you need, instead of exporting its JSON and importing it back. You can also export any panel as PNG or SVG to share with your team on Slack.

5. Improved dashboard header. Adding a variable is now inline. You create it from the dashboard itself. The header stays sticky, but its size is optimized. So you don't lose the context of your variables and time range, and you get a larger surface for the panels.

6. More control over panels. New panel styling options like chart fill mode and line interpolation are part of the schema, so you or an agent can set them directly.

Getting Started with SigNoz

Agent-native observability takes more than exposing your telemetry to an agent. It takes a data model designed for one. That is what the new dashboard schema does, and it is the direction the rest of SigNoz is heading, i.e., treating agents as first-class users.

The best way to start is SigNoz Cloud. You get 30 days of unrestricted access to every feature, and our team can help you onboard or migrate your existing setup.

If you prefer to keep your data in your own environment, SigNoz is also available as a self-hosted Enterprise edition (BYOC or on-premises) and as the open-source Community edition.