ObservabilityCON 2026
Grafana Labs, Pier 27 San Francisco, Oct 19–21
Table of Contents
1. Overview
Grafana Labs's annual observability-practitioner conference. Three days at Pier 27 on the Embarcadero: one optional workshop day (separate ticket, hosted at Hyatt Centric Fisherman's Wharf), two conference days. Framing for this edition is "AI for observability. Observability for AI. Solutions for all." — the pivot that carries the whole program is agentic tooling inside Grafana Cloud (Grafana Assistant, Agent Observability, Adaptive Telemetry), which shows up in the workshop titles before the keynote list is even published.
Distinguishing frame this year: the 2026 program treats AI workloads as the observed system (Agent Observability inspecting prompts, tool calls, and agent execution) and simultaneously as the observer (Grafana Assistant writing queries, summarising logs, driving dashboards). That symmetry is unusual — most vendor conferences pick a side — and it maps directly onto the agent-telemetry problem the wal.sh corpus is tracking.
Note that GrafanaCon 2026 (April 20–22, "Grafana 13") and ObservabilityCON 2026 (Oct 19–21) are separate events on the 2026 Grafana Labs calendar — the naming had drifted for a while, but the 2026 sites now list them as distinct.
1.1. Basics
| Field | Value |
|---|---|
| Event | ObservabilityCON 2026 |
| Dates | |
| Main venue | Pier 27, 27 The Embarcadero, San Francisco, CA 94111 |
| Workshop venue | Hyatt Centric Fisherman's Wharf San Francisco |
| Organizer | Grafana Labs |
| Tagline | "AI for observability. Observability for AI. Solutions for all." |
| URL | https://grafana.com/events/observabilitycon/ |
| Registration | Open; 30% early-bird discount (limited) |
| State | Workshop list published; main-track keynotes and sessions not yet announced (as of 2026-09-08) |
1.2. Sponsors
| Tier | Sponsor |
|---|---|
| Pioneer | AWS |
| Guru | Google Cloud |
| Champion | Crestdata, Gremlin |
AWS at Pioneer tier explains the presence of the Instrument and observe production-ready agentic apps with Grafana Cloud and AWS workshop (joint-branded lab, below).
2. Program shape
| Day | Time | Block |
|---|---|---|
| Mon Oct 19 | 08:30–17:30 | Hands-on workshops (Hyatt Centric Fisherman's Wharf; separate ticket) |
| Tue Oct 20 | 09:00–20:00 | Opening keynote, technical sessions, Expo Hall, networking, welcome party (Pier 27) |
| Wed Oct 21 | 09:00–17:30 | Technical sessions, Expo Hall, networking (Pier 27) |
3. Workshops (Mon Oct 19)
Five hands-on labs, all at the Hyatt Centric. Optional; separate ticket.
- Intro to observability for the AI era — Grafana Cloud observability workflows plus AI tools for issue identification, troubleshooting, and root-cause analysis. The on-ramp lab.
- Advanced Grafana: From dashboard to AI command center — power-user session on turning a dashboard into a "central hub that brings all your data together." The dashboards-as-control-plane pitch, made concrete.
- The power user's guide to the AI assistant in Grafana Cloud — advanced Grafana Assistant usage: writing queries, summarising logs. The one to watch for anyone assessing LLM-authored PromQL/LogQL quality.
- Instrument and observe production-ready agentic apps with Grafana Cloud and AWS — the joint-branded AWS lab. Guided labs to inspect "conversations, prompts, tool calls, and agent execution" — i.e. Agent Observability applied end-to-end.
- Building telemetry pipelines with Grafana Alloy — hands-on with Alloy, Fleet Management, and Assistant for telemetry collection, processing, and visualisation. The OTel-Collector-successor training.
4. Themes (advertised for 2026)
Pulled from the official page and workshop titles — the topic surface Grafana Labs is positioning for 2026:
- Agentic AI tools and workflows in Grafana Cloud
- Grafana Assistant (LLM-authored dashboards, queries, log summaries)
- Agent Observability (prompts, tool calls, agent execution as first-class telemetry signals)
- Adaptive Telemetry (cost control on ingested data)
- OpenTelemetry adoption and Alloy pipeline building
- Site reliability, digital experience monitoring, AI observability
- Scaling observability across teams and organisations
- Improving incident response and reliability
- Reducing costs while maintaining visibility
Framing is heavier on practitioner-lessons-from-production than on vendor roadmap — whether that survives contact with the actual keynote slot list is the open question. The main-track program is not published yet.
5. Announcements
The pre-event guess held: ObservabilityCON carried product announcements, all three for Grafana Cloud and all aimed at what Grafana calls autonomous operations. Agents take over recurring observation, evidence gathering and preparing changes, while people keep the decisions about production.
| Feature | Status | What it does |
|---|---|---|
| Assistant Watchers | GA | Always-on agents for conditions you define: baseline from historical telemetry, check signals periodically, report with evidence, and start an investigation when something critical turns up |
| Assistant Investigations | GA | Gather evidence about what changed across metrics, logs, traces and profiles, test hypotheses, and return findings with the source queries |
| Grafana Managed Sandboxes | Public preview | Controlled environments where an agent inspects repository code, prepares a fix from the investigation context and live telemetry, and runs checks before people review it |
Read as a pipeline, a watcher notices, an investigation explains, and a sandbox drafts the fix. Only the last step proposes a change, and it stops at review. The post does not mention MCP or use the phrase "AI SRE".
Keynote speakers and the main-track session list were never captured here; the post covers the launches only.
6. ObservabilityCON on the Road
One-day satellite events for people who can't make San Francisco. Four cities confirmed:
| City | Date |
|---|---|
| São Paulo | |
| London | |
| Madrid | |
| Bengaluru |
Structure of the satellites (talks vs. workshops, streamed content from SF vs. local program) is not yet documented; treat these as "same brand, different depth" until confirmed.
7. Alignment with wal.sh research
Three threads on this year's program land on active work.
Agent Observability as first-class telemetry. Workshop 4 (Grafana Cloud
- AWS) treats prompts, tool calls, and agent execution as observability
signals in their own right. That is the same position the Agent Telemetry Systems note takes from the research side, and the same shape crowsnest adopts locally for LLM traffic. If a main-track talk lands in this territory, it goes on the crowsnest reading list.
LLM-authored queries and dashboards. Workshop 3 (power user's guide to the Grafana Assistant) is the vendor version of the generate freely, check mechanically position — generator on one side, PromQL/LogQL correctness on the other. The interesting question is what the checker is: a syntactic parse, a dry-run against sample data, or something closer to a schema-pact contract.
Cost as a first-class concern. Adaptive Telemetry as a theme sits alongside "reducing costs while maintaining visibility" as a program axis — observability's data-gravity problem, publicly acknowledged. The compression-vs.-fidelity trade-off is the same one that shows up in log-retention decisions across the corpus.
8. Related
- GrafanaCon 2025 — the previous year's Grafana Labs conference (Seattle, May 6–8, same audience shape).
- GrafanaCon 2026 (Apr 20–22, "Grafana 13" reveal) — separate event, no note yet; Raj Dutt + Torkel Ödegaard on the opening keynote.
- Observability Summit NA 2026 — parallel observability event, same practitioner audience, different organiser.
- Agent Telemetry Systems — AI observability from the research side.
- REPL-Driven Ticket Triage — incident triage as data an agent queries; Watchers that start Investigations are the hosted form of that loop.
- Agent Permission Guardrails — where Managed Sandboxes sit: the agent prepares the change, and approval stays with people.
- crowsnest — the site's own local LLM observability tool.
- AI Tinkerers Boston 2026-06-29 — Poliakov's DBOS talk on durable and observable AI agents; closest adjacent recent talk.
9. Sources
- grafana.com/events/observabilitycon/ — official conference page.
- grafana.com/events/observabilitycon/agenda/ — Monday workshop list.
- grafana.com/events/grafanacon/ — sister event (April 20–22), used to confirm the two conferences are distinct.
- Grafana Labs interest-list email, 2026-07-10 (
update@grafana.com). - grafana.com/blog/grafana-ai-autonomous-operations/ — post-event announcements (fetched 2026-10-22).
10. Open questions / TODO
[X]Confirm dates to October 19–21, 2026.[X]Confirm venue to Pier 27, The Embarcadero, San Francisco.[X]Confirm workshop venue to Hyatt Centric Fisherman's Wharf.[X]Confirm On-the-Road satellite dates to SP Nov 4, London Nov 5, Madrid Nov 24, Bengaluru Dec 8.[X]GrafanaCon vs. ObservabilityCON to distinct 2026 events (GrafanaCon Apr 20–22, ObservabilityCON Oct 19–21).[X]Announcements — Assistant Watchers and Investigations (GA), Managed Sandboxes (public preview); see Announcements (2026-10-22).[ ]Keynote speakers — not announced as of 2026-09-08; not captured since.[ ]Main-track session list — not announced (Mon workshops only).[ ]Satellite format — talks vs. workshops, streamed vs. local program, per-city agendas.[ ]Sponsor list beyond Pioneer/Guru/Champion tiers.[ ]CFP status / whether a Community track exists.[ ]Post-event: pull per-talk notes into anincludes/split.