---
title: "Manage Grafana Assistant pricing, usage, and limits | Grafana Cloud documentation"
description: "Understand Grafana Assistant pricing and usage, monitor usage, and manage limits and access for your organization."
---

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

# Manage Grafana Assistant pricing, usage, and limits

Grafana Assistant is billed based on usage. The pricing model covers active AI users, token usage, Investigations, and what’s included with your plan. For details, refer to [Grafana Assistant pricing](https://grafana.com/docs/grafana-cloud/platform/pricing-and-usage/assistant/). For current rates, refer to the [Grafana pricing page](https://grafana.com/pricing/).

This document explains the terms behind Assistant usage, how each capability is billed, the usage you see in the product, and how to monitor usage, set limits, and control access for your organization.

Usage, limits, and billing apply to Grafana Assistant in Grafana Cloud. Anyone who uses Assistant sees the in-product limits described in this document. Monitoring usage, setting limits, and controlling access require administrator permissions.

## Before you begin

Before you manage Assistant usage and limits, confirm the following:

- You can open Assistant and Investigations in your stack.
- You have the **Assistant Admin** role, or the `grafana-assistant-app.usage:read` and `grafana-assistant-app.usage:write` permissions, to view usage and configure limits.
- You know your Grafana Labs account contact for billing and usage questions.

## Key terms

Assistant pricing and limits use the following terms.

Expand table

| Term                    | What it means                                                                                                                                                             |
|-------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Token                   | The unit Assistant usage is measured in. Both the text you send and the text the model generates consume tokens.                                                          |
| Active AI user          | A user who uses Assistant during a calendar month, shown as **Active Users** in the product. Active users count toward your monthly active user limit.                    |
| Conversation            | A single chat with Assistant. Starting a new chat starts a new conversation.                                                                                              |
| Context window          | The amount of a conversation the model can consider at once.                                                                                                              |
| Compaction              | Assistant condensing an ongoing conversation into a summary as it nears the context window.                                                                               |
| System-initiated tokens | Tokens consumed by work Assistant starts on its own, such as Watcher runs, drawn from a shared organization pool.                                                         |
| Service account         | A non-human Grafana identity Assistant uses to run system-initiated work. System-initiated usage counts toward service-account token limits.                              |
| Usage limit             | A monthly cap on usage. Limits apply to active users, investigations, and tokens, with token limits at the stack, per-user, team, and system-initiated token pool scopes. |

## Understand how Assistant capabilities are billed

Grafana Assistant includes the following capabilities. All of them run on tokens, and usage above the allowances included with your plan is billed as AI tokens.

Expand table

| Capability       | What it does                                                                                                                         | How it’s billed                                                                                                                                                                                                                                                                                                                                                                                                        | Learn more                                                                                                                                                                                                     |
|------------------|--------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| Assistant chat   | Conversations with Assistant in Grafana, Workspace, Slack, Microsoft Teams, the `gcx` CLI, and through the Grafana Cloud MCP server. | Messages and Assistant actions count toward your tokens and make you an active AI user on every surface. In `gcx`, only `gcx assistant` commands count. Connecting through the Grafana Cloud MCP server counts you as active even if you use no tokens, and only Assistant features consume tokens.                                                                                                                    | [Assistant Workspace](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/workspace), [gcx CLI](https://grafana.com/docs/grafana/latest/as-code/observability-as-code/grafana-cli/gcx/) |
| Investigations   | Multi-step root cause analysis across your telemetry, started manually or by an alert, IRM webhook, or Watcher.                      | Billed in tokens only, with no separate fee per investigation. Manual investigations count toward your usage and make you an active AI user; automatically triggered investigations draw from the system-initiated token pool and don’t count anyone as an active user. Prompts you send inside an automated investigation count toward your own usage. Metering started October 1, 2026, with no retroactive billing. | [Investigations](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/investigation)                                                                                                     |
| Watchers         | Always-on agents that check your telemetry on a schedule and report notable changes.                                                 | Runs, calibration, and automatic recalibration reviews use system-initiated tokens and don’t count anyone as an active AI user. Investigations a Watcher launches are billed as investigations.                                                                                                                                                                                                                        | [Use Watchers](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/watchers)                                                                                                            |
| Automations      | Saved prompts that run on demand, on a schedule, or from an event.                                                                   | Runs with **Personal access** count toward a user: the invoker for manual runs, the creator for scheduled runs, and the identity the trigger supplies for event-triggered runs, such as the author of the Slack message. Runs with **Organization access**, including every GitHub-triggered run, use the system-initiated token pool.                                                                                 | [Use Automations](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/automations)                                                                                                      |
| Pages            | Interactive pages built from your Grafana data in conversation, which you refine and share. Public preview.                          | Page creation and refinement happen in Assistant conversations, which use tokens.                                                                                                                                                                                                                                                                                                                                      | [Grafana Pages](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/pages)                                                                                                              |
| Assistant Search | Finds indexed Grafana resources and Assistant memory from a description.                                                             | Using it on its own doesn’t make you an active AI user, but the tokens it consumes count toward your token usage.                                                                                                                                                                                                                                                                                                      | [Use Assistant Search](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/assistant-search)                                                                                            |
| Skills           | Reusable instructions and runbooks Assistant uses when relevant to a request.                                                        | No separate charge. Skill content adds to the tokens of the conversation or run that uses it.                                                                                                                                                                                                                                                                                                                          | [Create Skills](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/skills)                                                                                                             |
| Rules            | Always-on instructions applied to every conversation.                                                                                | No separate charge. Rule content adds to the tokens of the conversations it applies to.                                                                                                                                                                                                                                                                                                                                | [Configure Assistant rules](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/configure/rules)                                                                                                 |

Using Assistant or any of these capabilities counts toward your token usage. For where to review usage by user, capability, and service account, refer to [Monitor usage with Usage Analytics](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/pricing/#monitor-usage-with-usage-analytics).

### Review Watcher usage

Watcher runs draw from your organization’s shared system-initiated token pool and count toward your stack and shared service-account token limits. Creating more Watchers doesn’t increase your allowance.

Open a Watcher’s detail page under **AI** &gt; **Watchers** to review its estimated token consumption, the applicable monthly limit, and the reset date. The estimate doesn’t include investigations the Watcher launches, which are billed as investigations. When that limit is reached, Watcher runs and calibration stop until it resets. Scheduled Watchers stay active and resume automatically after the reset.

> Note
> 
> Watcher usage is free until December 1, 2026. From December 1, Watcher usage is billed as part of your Assistant token usage.

For setup instructions, refer to [Use Watchers](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/watchers).

## Limits you encounter in the product

Assistant shows two kinds of limits in the chat: a per-conversation context limit that keeps a single chat workable, and your organization’s monthly usage limits. The context limit doesn’t affect billing or your monthly usage.

Each conversation includes a context limit indicator that shows how much of the chat fits in the current context window. As a conversation nears the context window, Assistant may compact it. If a request still exceeds the window, Assistant may offer to start a new chat with a summary.

As monthly usage nears a configured limit, Assistant shows a warning in the chat input with your current usage. When a limit is exhausted, Assistant shows a limit message and blocks further use for that limit until it resets at the start of the next month or you change it. The limit message offers different actions depending on your permissions:

Expand table

| Your access                  | Action in the limit message                            |
|------------------------------|--------------------------------------------------------|
| Usage permissions            | Open **AI** &gt; **Usage** to review usage and limits. |
| Administrator on a Free plan | Option to upgrade for token limits.                    |
| No usage permissions         | Prompt to contact an administrator.                    |

Investigation usage warnings appear only in **Investigation** mode, because investigation usage applies only to investigation workflows.

## Monitor usage with Usage Analytics

The Usage Analytics dashboard shows how your organization consumes Assistant in the current stack.

Navigate to **AI** &gt; **Usage** to open the dashboard. It shows active users, user messages, investigations, and total tokens used for your stack, along with your current usage against token and investigation limits. Access requires the `grafana-assistant-app.usage:read` permission, which the **Assistant Admin** role includes by default. That role is granted to the Grafana organization **Admin** basic role.

Filter the dashboard by time range to review adoption and usage trends. Choose **Month to date**, **Last 7 days**, **Last 30 days**, or **Last 90 days**. The **Tokens** view breaks token usage down by capability, such as investigations and automations.

In the **Investigations** view, the investigation breakdown shows investigations per day and in a table. Most investigations start from alerts, and a few service accounts create them, so use **Group by** to see which teams, integrations, and alerts drive them. Group by **Team**, **Integration** for the IRM integration, **Alert** for the alert name, **Source** for how the investigation started, or **Creator**. Each view also shows the tokens that each group’s investigations used, including investigations that webhooks and alerts create. Under **Show**, select **Tokens** to rank the groups, the shares, and the chart by tokens instead of by investigation count. Investigations that started before Assistant began to record tokens for each investigation show no tokens. By default, the team view counts each investigation once, under its first team, which is usually the team from the alert or the IRM integration. Select **Every team** to count an investigation under every team that has access to it. Click a team to open its investigations. **No team** collects investigations without a team. To assign alert investigations to a team, refer to [Give a team access to alert investigations](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/configure/investigation-alerts/#give-a-team-access-to-alert-investigations). A **No Grafana team found** badge means that no Grafana team you can see has that name. If no team has the name, it gives nobody access to the investigations.

To see usage across your plan, broken down by individual user and service account, and to compare usage included with your plan against additional usage, select **Assistant** in the **Usage** tab of the **Cost Management and Billing** app. For more information, refer to [Understand Grafana Assistant usage and cost](https://grafana.com/docs/grafana-cloud/platform/cost-management-and-billing/understand-usage-cost/assistant/).

If you don’t have access to Usage Analytics, ask your Grafana Labs contact for a usage summary.

## Manage access and usage limits

You control who can use Assistant and how much they can consume. Set monthly user and token limits, or use RBAC to restrict access to specific users and teams.

### Set usage limits

Before you save usage limits, accept the Assistant terms on each stack you configure. Acceptance on one stack doesn’t apply to others. Eligible Grafana Cloud Free and Pro stacks accept terms automatically under your existing agreement. For instructions, refer to [Set up Grafana Assistant in Grafana Cloud](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/get-started/grafana-cloud).

Assistant supports the following limit scopes:

Expand table

| Limit scope                 | Applies to                                                 | Where to set it                                                    |
|-----------------------------|------------------------------------------------------------|--------------------------------------------------------------------|
| Active users                | Unique users per month across the stack                    | **AI** &gt; **Usage** &gt; **Active Users** &gt; **Edit limits**   |
| Stack                       | Total token usage across the stack                         | **AI** &gt; **Usage** &gt; **Tokens** &gt; **Edit limits**         |
| Per-user                    | Each user’s token usage, by default                        | **AI** &gt; **Usage** &gt; **Tokens** &gt; **Edit limits**         |
| Team                        | Per-user token usage for members of a team (Grafana Cloud) | **AI** &gt; **Usage** &gt; **Tokens** &gt; **Edit limits**         |
| System-initiated token pool | Tokens for work Assistant starts on its own                | **AI** &gt; **Usage** &gt; **Tokens** &gt; **Edit limits**         |
| Investigations              | Investigations started per month across the stack          | **AI** &gt; **Usage** &gt; **Investigations** &gt; **Edit limits** |

Setting and changing limits requires the `grafana-assistant-app.usage:write` permission, which the **Assistant Admin** role includes by default. Click **Save** to apply your changes.

Keep these rules in mind:

- Set the monthly active user limit to `0` to disable the cap and allow unlimited active users. A positive number tracks monthly active users against the cap, shown in the **Active Users** view.
- Enter token limits in millions of tokens.
- A team token limit overrides the per-user default for that team’s members.
- If a user belongs to multiple teams with token limits, the highest per-user team limit applies.
- The stack limit applies to total usage across the stack, regardless of per-user, team, or system-initiated token pool limits.
- Token limits apply to month-to-date usage, including usage from before you set the limit. If a user already exceeds a new limit, Assistant blocks that user until the next monthly reset, or until you raise or remove the limit.

### Increase token limits during an incident

In Grafana Cloud, you can prepare a team with a higher per-user token limit so responders can keep using Assistant after they reach their usual per-user limit. This raises their allowance; it doesn’t reserve tokens. Usage follows your plan’s normal billing.

You need the `grafana-assistant-app.usage:write` permission, which the **Assistant Admin** role includes by default, and permission to manage team membership.

Before an incident:

1. Create a dedicated Grafana team for temporary access.
2. In **AI** &gt; **Usage**, open the **Tokens** view and set a higher per-user limit for the team. Choose unlimited only if your stack allows it.
3. Decide who manages temporary membership and when they remove it. If you use an identity provider, make sure its group membership syncs to the Grafana team.
4. Check that your stack has enough remaining token allowance.

During an incident, add responders to the team. Membership changes can take time to propagate, so confirm the higher limit applies before you rely on it. If a user belongs to multiple teams with limits, the highest team limit applies.

A team override changes only the per-user limit. If another limit blocks usage, adjust that limit too:

Expand table

| Limit reached                | Action                                                                                         |
|------------------------------|------------------------------------------------------------------------------------------------|
| Per-user limit               | Add the user to a team with a higher limit, or raise their applicable limit.                   |
| System-initiated token limit | Raise the shared system-initiated token limit. Team overrides don’t apply to service accounts. |
| Stack limit                  | Raise the stack limit. Higher user or system-initiated token limits don’t bypass it.           |

You can only increase limits within the maximums allowed for your stack. Other limits, including investigation limits, still apply.

After the incident, remove temporary memberships and restore any limits you changed. Assistant doesn’t expire these changes automatically. Tokens already consumed still count toward monthly usage. If usage exceeds a restored limit, that limit blocks further use until the monthly reset or another increase.

For API instructions, refer to [Token limits API](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/reference/http-apis/token-limits).

### Restrict access with RBAC

Use role-based access control (RBAC) to restrict Assistant access to specific users or teams. You can modify custom roles to include or exclude Assistant permissions. Refer to [Manage Assistant access with RBAC](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/privacy-and-security/rbac) for the Assistant roles and permissions.

To remove Assistant access from a basic role such as Viewer or Editor, update the basic role with the RBAC API and remove the Assistant permissions you don’t want that role to grant. Refer to [Manage Grafana RBAC roles](https://grafana.com/docs/grafana/latest/administration/roles-and-permissions/access-control/manage-rbac-roles/#update-basic-role-using-the-rbac-api) for API instructions.

Changing basic role definitions may affect other user permissions. If you need help with RBAC configuration, contact Grafana Labs support to review your requirements.

## Next steps

- [Grafana Assistant pricing](https://grafana.com/docs/grafana-cloud/platform/pricing-and-usage/assistant/)
- [Grafana pricing page](https://grafana.com/pricing/)
- [Use Grafana Cloud MCP tools](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/configure/cloud-mcp)
- [Investigations](https://grafana.com/docs/grafana-cloud/platform/grafana-assistant/platform/investigation)
