Grafana Assistant
by Grafana Labs
AI observability agent in Grafana Cloud that writes PromQL, LogQL and TraceQL from plain English, investigates incidents and proposes code fixes as pull requests
Grafana Assistant is the AI agent built into Grafana Cloud that answers plain-language questions about production telemetry, builds dashboards and queries, and runs incident investigations. It is for SRE, platform and developer teams already on Grafana who want faster root-cause analysis without every engineer knowing PromQL or LogQL.
Grafana Assistant is Grafana Labs' AI agent for observability, running inside Grafana Cloud and available to self-managed Grafana OSS and Enterprise through a one-click connection to the Grafana Cloud backend. It entered public preview in August 2025 and reached general availability at ObservabilityCON on October 8, 2025, when it gained PromQL, LogQL and TraceQL generation from plain-language prompts and Assistant-specific roles built on Grafana Cloud RBAC. In July 2026 Grafana made Assistant Investigations (an autonomous incident-response agent that correlates signals, forms hypotheses and recommends remediation), Workspace, Automations, the Grafana Cloud MCP server and the gcx CLI generally available. At ObservabilityCON on October 20, 2026 it added Assistant Watchers (GA), always-on agents that learn normal telemetry patterns from a plain-language description and route evidence-backed findings through Grafana Alerting, plus Managed Sandboxes in public preview, where Assistant clones a repository, checks code against production behaviour and opens a pull request for human review; Grafana says it never merges or deploys. Assistant works from Grafana Cloud, Slack, Microsoft Teams, IDEs, MCP clients, the API and gcx. Billing has two dimensions, active AI users and tokens: the free tier covers 3 active AI users with 40M tokens each, Pro is $20 per active AI user with additional tokens from $2 per 1M, and Investigations metering began October 1, 2026. SailPoint is a named user. An IDC study sponsored by Grafana reported 63% lower MTTR across four organisations, which IDC itself calls directional. It competes with Datadog Bits AI, Dynatrace Davis and New Relic AI.
Heads of SRE or platform engineering at companies already running Grafana Cloud (or self-managed Grafana willing to connect to it) who want to shorten incident investigation and widen who can query telemetry.
Engineers ask questions in plain language and get investigations, queries and proposed fixes drawn from their own metrics, logs and traces, without leaving Grafana or Slack.
At a Glance
- Category
- Infrastructure & Cloud
- Pricing
- Freemium, Usage-based
- Target Market
- CTOs, VPs of Engineering, SRE Teams, Platform Engineers, DevOps Engineers
- Deployment
- Cloud-first, Hybrid
- Customers
- Grafana Cloud: more than 10,000 organisations (company figure, October 2026); Assistant-specific count not published
Key Features
- ✓Natural-language querying
Generates and refines PromQL, LogQL and TraceQL from plain-English prompts, so non-specialists can query metrics, logs and traces.
- ✓Assistant Investigations
An autonomous incident-response agent that starts from an alert or incident, correlates signals, forms hypotheses and recommends mitigation; generally available since July 2026.
- ✓Assistant Watchers
Always-on agents, GA since October 2026, that learn normal behaviour from history and surface evidence-backed changes through Grafana Alerting.
- ✓Managed Sandboxes (preview)
Lets Assistant clone a GitHub repository, compare code with live telemetry, prepare a fix and open a pull request for human approval.
- ✓Agent Observability
Tracks what AI agents do and what they cost using OpenTelemetry, shown alongside logs, metrics and traces for other services.
- ✓Assistant RBAC and controls
Assistant-specific roles on top of Grafana Cloud RBAC let admins control access to rules, investigations, integrations and MCP server management.
- ✓Works where engineers already are
Available in Grafana Cloud, Slack, Microsoft Teams, IDEs, MCP clients, the API and the gcx command-line tool.
Capabilities
Use Cases
- •Alert-triggered root-cause analysis
An alert fires and an Investigation automatically correlates metrics, logs and traces, returning findings and a suggested mitigation to the on-call engineer.
- •Self-service telemetry for developers
A product engineer asks in plain English why latency rose and gets a generated query and dashboard, without waiting for an SRE.
- •Catching unknown unknowns
A Watcher told to monitor checkout errors learns the baseline and flags an unusual change that no threshold rule covered.
- •From incident to pull request
After an investigation, a Managed Sandbox drafts a code fix against the real repository and opens a pull request for review before merge.
- •Faster log spike triage
SailPoint's DevOps team reports cutting the time to find log spikes from 15 to 20 minutes down to about five by asking Grafana directly.
Ideal For
Best For
- ✓SRE teams that want automated root-cause investigations triggered from alerts or incidents
- ✓Developers and product engineers who need telemetry answers without learning PromQL, LogQL or TraceQL
- ✓Platform teams watching for anomalies that fixed-threshold alert rules miss (Assistant Watchers)
- ✓Teams that want an agent to draft a code fix from production evidence and open a reviewed pull request
- ✓Organisations standardising on OpenTelemetry that want agent activity and cost observed next to other services
Not Ideal For
- ✗Air-gapped or strictly on-premise environments: the AI features require a Grafana Cloud connection, so telemetry and queries leave the customer's perimeter
- ✗Teams not on Grafana; the Assistant only works over Grafana Cloud or Grafana instances connected to it
- ✗Budget owners who need fixed costs, since token usage from automated investigations and Watchers (billed from December 1, 2026) is variable
Deployment
Market & Ratings
Grafana Cloud: more than 10,000 organisations (company figure, October 2026); Assistant-specific count not published
Market Analysis
Pros
- ✓Generally available since October 2025 with a steady release cadence (Investigations GA July 2026, Watchers GA October 2026)
- ✓Transparent, published pricing with a usable free tier
- ✓Runs where engineers already work: Slack, Teams, IDEs, MCP clients and a CLI
- ✓Human approval stays in the loop for remediation; proposed fixes arrive as pull requests
Cons
- ✗Requires Grafana Cloud; there is no air-gapped or fully on-premise AI option, so telemetry and queries leave the customer's VPC (OpsTree comparison, written by a competitor)
- ✗Outcome claims rest on vendor-sponsored evidence: IDC's 63% MTTR figure comes from four organisations and IDC calls it directional; Let's Data Science notes the October 2026 launch is not independent proof of better incidents
- ✗Token-metered investigations and Watchers make monthly cost hard to predict, and free-tier limits block new chats until the next billing period
- ✗Still needs query literacy for deep work; it helps write PromQL/LogQL but does not replace that skill (OpsTree)
Pricing
Free
$0
- ✓3 active AI users/mo
- ✓40M tokens per user/mo
- ✓25M tokens/mo for service accounts
- ✓Hard limits, community support
Pro
From $20 per active AI user/mo
- ✓40M tokens per user included
- ✓Additional tokens from $2 per 1M
- ✓$19/mo platform fee includes 3 active AI users
- ✓8x5 email support
Enterprise
Contact for pricing
- ✓Minimum commit $25,000/yr
- ✓Premium support
- ✓Public Cloud, Federal Cloud or Bring Your Own Cloud
Grafana publishes list prices: Assistant is billed per monthly active AI user ($20 on Pro, 40M tokens each) plus tokens above the allowance from $2 per 1M. Investigations have no per-run fee but their tokens are metered from October 1, 2026, Watcher usage becomes billable token usage from December 1, 2026, and automated runs draw on a shared system-initiated token pool. Free-tier limits are hard stops. Enterprise needs a $25,000/yr minimum commit and contracts can set different rates.
Security & Compliance
Connect
Sources
This page was written from 9 sources, 5 on domains other than grafana.com.
- 1.grafana.com — ai assistantvendor
- 2.grafana.com — assistantvendor
- 3.grafana.com — assistantvendor
- 4.grafana.com — observabilitycon 2026 ai updatesvendor
- 5.gilbane.com — grafana labs releases grafana assistant
- 6.itweb.co.za — mYZRXv9gn2eMOgA8
- 7.letsdatascience.com — grafana adds agent observability and assistant watchers at o
- 8.opstree.com — opstree.com
- 9.hn.algolia.com — search
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Ollama
Run open-weight LLMs locally with one command, then burst to Ollama's cloud on the same API
Inferact Platform
Managed vLLM inference from the creators of vLLM, on your own GPUs or on Inferact Cloud
Perplexity SPACE
Open-source Firecracker microVM sandbox for AI agents, the Apache 2.0 runtime behind Perplexity Computer
Gimlet Cloud
Multi-silicon inference cloud that splits AI agent workloads across GPUs, CPUs and accelerators