---
title: "Understand Grafana Agent Observability pricing | Grafana Cloud documentation"
description: "Learn how Agent Observability pricing works, what is billed, and where to find current rates."
---

> For a curated documentation index, see [llms.txt](/llms.txt). For the complete documentation index, see [llms-full.txt](/llms-full.txt).

# Understand Grafana Agent Observability pricing

Grafana Agent Observability is billed based on usage. This article explains how pricing works, defines billing terms, and shows you where to find current rates and included usage.

Pricing details and plan tiers change over time. Refer to the [Grafana pricing page](/pricing/) for the latest rates and what each plan includes. Different plan types offer different pricing structures and included usage.

## Before you begin

- Make sure Agent Observability is enabled on your Grafana Cloud stack.
- Make sure your applications are sending generation data.
- Contact your Grafana Labs account team for billing and usage questions.

## Understand what is billed

Agent Observability uses different billing units for different data and feature types. Generation usage is billed separately from token usage for LLM-based evaluations and guards.

OpenTelemetry metrics and traces generated by Agent Observability SDKs use the existing Grafana Cloud telemetry pricing for those data types. They aren’t included in Agent Observability generation pricing. Refer to the Grafana pricing page for current telemetry rates and included usage.

## Know the pricing model

The pricing model uses different billing units for generation data, LLM-based evaluations and guards, and telemetry. Grafana Cloud offers different plan types for Agent Observability. Each plan type has different pricing structures and included usage allowances. Refer to the [Grafana pricing page](/pricing/) for current plan details and what’s included in each tier.

### Generations

A generation is one call to an AI model made by your monitored agent or application. Each time your application calls an LLM provider and the SDK exports that interaction, Agent Observability counts it as a generation.

One agent task can produce multiple generations if it makes multiple model calls. Generation billing applies to ingesting and storing this telemetry. It doesn’t cover the cost of running your model.

### Evaluation and guard usage

Evaluations and guards can use different evaluator types. Usage depends on whether the evaluator calls an LLM.

LLM judge evaluations and LLM-based guards use token-based usage because Grafana runs an LLM to evaluate or guard the interaction. This token usage counts toward the **system-initiated token pool** for your stack. Deterministic evaluators and guards, such as regex, JSON schema, heuristic, redact, and tool filter rules, don’t use LLM tokens unless they call an LLM-backed evaluator.

### Telemetry usage

Agent Observability SDKs can emit OpenTelemetry metrics and traces alongside generation data. These metrics and traces help you monitor performance, errors, latency, and token usage in your existing Grafana Cloud observability tools.

Telemetry usage follows the standard Grafana Cloud pricing and limits for metrics and traces. Agent Observability doesn’t add a separate usage cap for this telemetry.

## View usage analytics

Use Agent Observability analytics views to monitor generation volume, latency, errors, token usage, and cost patterns. This visibility helps you understand adoption, identify expensive workflows, and manage usage.

## Billing

Pricing and metering for generations and evaluation and guard token usage start on October 1, 2026. Telemetry billing follows standard Grafana Cloud pricing and is already active. Refer to the [Grafana pricing page](/pricing/) for current rates and included usage.

## Next steps

- [Grafana pricing page](/pricing/)
- [Introduction to Agent Observability](/docs/grafana-cloud/observe-and-act/agent-observability/introduction)
- [Optimize cost and performance](/docs/grafana-cloud/observe-and-act/agent-observability/guides/cost-optimization)