Databricks Launches Always-on Genie: What the Autonomous Agent Costs, What’s GA, and the 30 November MCP Deadline

On 11 November 2026 Databricks announced always-on Genie and GA for Genie Ontology. Always-on is still preview/beta. Named-user LLM usage is free until 31 January 2027, then 150 DBUs ($10.50 in US East) and $0.070/DBU, plus compute. The beta Genie MCP endpoint ends 30 November 2026.

Databricks Launches Always-on Genie: What the Autonomous Agent Costs, What's GA, and the 30 November MCP Deadline

Short answer: On 11 November 2026 Databricks announced a second-generation, always-on Genie that can run in the background, delegate to sub-agents, and bring decisions to a priority inbox and email. Genie Ontology is generally available. Always-on Genie and Ontology Manager are coming soon to preview and beta, per the press release (the same-day Ontology blog also calls Genie Search generally available; the press FAQ still lists Search as coming soon). Pricing, checked 12 November 2026 on AWS US East: Genie One and Genie Agents LLM usage is free for named users until 31 January 2027, then 150 free DBUs per user per month and $0.070 per DBU after that (150 DBUs = $10.50 in US East). SQL warehouse compute is billed separately. The beta Genie MCP endpoint is retired on 30 November 2026.

Checked 12 November 2026 (Asia/Tehran) against Databricks’ press release, Ontology blog, Genie pricing page, cost and MCP docs, and n8n’s Genie One MCP documentation. We have not run Genie ourselves. Vendor benchmarks and customer quotes are labelled as claims.

Databricks wants Genie to stop waiting for a prompt. In its 11 November 2026 press release, the company describes a second generation of Genie as an always-on enterprise agent: you give it a goal and boundaries, and it keeps the work moving, delegates to sub-agents, follows up on missing inputs, and pings you when a person has to decide. That is a different product from a SQL chatbot, and the pricing page is finally explicit about what the promo covers and what it does not.

The catch, which the same press release states plainly, is that the new always-on behaviour is not generally available yet. Genie and Genie Ontology are GA. Always-on Genie, Ontology Manager and — in the press FAQ — Genie Search are coming soon to preview and beta. If you already use Genie from n8n or another MCP client, the more urgent date is 30 November 2026, when Databricks turns off the beta Genie MCP endpoint.

What Databricks announced

ItemWhat it isAvailability (Databricks, 11 Nov 2026)
Always-on GenieBackground agent: sub-agents, priority inbox, email alerts, own computeComing soon to preview and beta (press release)
Genie OntologyAgentic ontology of definitions, metrics, people and processesGenerally available (press release and Ontology blog)
Genie SearchOne governed index across the lakehouse and connected appsPress FAQ: coming soon to preview/beta. Ontology blog and Search docs (updated 9 Nov): generally available. Not reconciled
Ontology ManagerUI to curate and govern business contextComing soon in beta (press and Ontology blog agree)
Genie MCP / pluginsGenie One as an MCP server; plugins for ChatGPT, Claude and Grok Bot/CursorGenie One MCP is generally available via Unity Gateway; beta URL ends 30 Nov 2026
Live artifactsData-driven slides, dashboards and reports that stay currentAnnounced; no separate GA date on the pages we checked

Sources: the press release and its built-in FAQ, the Genie Ontology GA blog, and the Genie One MCP docs (updated 11 November 2026).

What always-on Genie is supposed to do, once you can turn it on: run in the background on its own compute, take daily work (email, Slack, meeting prep) or longer jobs (a quarterly forecast, usage monitoring, a weekly revenue brief), and surface only the decisions. A new inbox lists items that need attention. Each alert arrives as its own unread email, with later updates on the same thread and a link back into the Genie app. Databricks says every action stays inside Unity Catalog permissions and Unity Gateway approval rules and budgets, and is auditable.

Genie Search, when you have it, is meant to index knowledge where it already lives — the press names Google Drive, Jira and Confluence, and Databricks talks about 100+ connectors that can ingest into the lakehouse, index in place, or federate live through MCP. The Search setup docs (updated 9 November 2026) list Google Drive, Jira, Confluence and Workday as Search sources, with MCP connectors for live lookups and write actions (for example Gmail or Slack). The Ontology blog adds SharePoint in beta. Account admins connect Search sources once; MCP connectors are signed in per user.

What it costs

This is what the Genie pricing page showed on 12 November 2026, with the plan toggle on Premium / AWS and a US East rate. Databricks notes that the dollar value of a DBU varies by region, and that Azure and Google Cloud have their own selectors.

ProductPromo (until 31 January 2027)After the promo (from 1 February 2027, unless noted)
Genie One (chat, search and the AI coworker for business users)LLM usage free for named users150 free DBUs per user per month, then $0.070 / DBU, plus compute
Genie Agents (curated, no-code agents)LLM usage free for named usersSame 150 DBU allowance and $0.070 / DBU, plus compute
Genie Code (coding assistant for data engineers and scientists)25% off the $0.070 / DBU list price on billed usage; 150 free DBUs already apply150 free DBUs per user per month, then $0.070 / DBU, plus compute
Genie OntologyLLM usage free“Price coming soon”

Footnotes on the pricing page: (1) free usage applies to Genie LLM usage only; Genie compute (for example SQL Serverless) is billed separately. (2) 150 DBUs = $10.50 in US East.

The cost-monitoring docs and consumption guide match that page: until 31 January 2027, Genie One and Genie Agents usage by users is free and does not count toward the 150 DBU allowance or toward Genie budgets; Genie Code is billed beyond the allowance; service principals get no free allowance and are billed from the first DBU; the 25% Code discount is applied in DBU metering. Genie budgets can alert or block pay-as-you-go LLM usage, with a maximum of 1,000 budgets per account, four shared thresholds and 100 per-user overrides per budget. Budgets cannot remove a user’s free monthly usage, and they do not cover SQL warehouse or other compute.

Our own arithmetic, using the US East list rate of $0.070 / DBU, after the promo ends, and excluding compute:

  • A named user who stays under 150 DBUs of Genie One or Agents: $0 in Genie LLM fees that month.
  • 500 billed DBUs above the allowance: 350 × $0.070 = $24.50.
  • 10 users, each 150 DBUs: $0 in LLM fees ($105 of list value is included).
  • During the promo, the same Genie One or Agents usage by named users is $0 in LLM fees; a service principal running the same work would be billed.

Databricks does not publish a DBU-per-question or DBU-per-always-on-job figure on these pages, so we have not invented a monthly seat equivalent. The consumption guide says usage varies with how often someone uses Genie, which products they use, task complexity, and the quality of available context.

Adoption and quality claims (vendor and customers)

Treat these as Databricks’ and its customers’ statements, not our measurements. The press release says more than 20,000 organizations worldwide, including 70% of the Fortune 500, rely on the Databricks Data + AI Platform. The Ontology blog says thousands of customers run Genie Agents in production, naming 7-Eleven, Virgin Atlantic and General Motors. A benchmark “built with Turing” is quoted as: Genie with Genie Ontology passed 91% of enterprise task runs, 1.6× the rate of “the strongest frontier agent”, while using less than half the tokens. Customer quotes in the press release include Headspace on self-service analytics, Slack on Genie Agents in Slack, and HelloFresh on a 75% reduction in time to identify ad winners on Meta, attributed to Genie and Unity Gateway. Ali Ghodsi, Databricks’ CEO, is quoted saying Genie has made him “a 10x more productive CEO”. None of these figures were independently checked by us.

What it means if you automate with n8n

Genie and n8n do different jobs. Genie is a governed assistant and agent runtime inside a Databricks workspace, priced in DBUs plus compute. n8n runs background workflows priced by executions, with unlimited users. n8n’s pricing page today (EUR, checked 12 November 2026) lists Cloud Starter at €20 a month billed annually (€24 billed monthly) for 2,500 executions and Pro at €50 (€60 monthly) for 10,000, still with unlimited users. A self-hosted Business plan is listed at €667 a month billed annually (€800 monthly) for 40,000 executions; it is not on n8n Cloud yet.

If your company already lives in Databricks, a practical split looks like this:

  • Ask Genie the data question, let n8n do the side effects. n8n’s Databricks Genie One MCP tile connects an n8n AI agent to Genie One through the workspace AI Gateway. Genie routes the question, writes and runs SQL, and returns an answer with an “Explore in Databricks” link. n8n can then write the result to Slack, a CRM or a ticket. The tile uses a Databricks OAuth2 credential with user login only. Required privileges, per n8n, include CAN RUN on each Genie Agent, CAN USE on at least one SQL warehouse, and EXECUTE on system.ai.genie_one_mcp.
  • Fix the MCP endpoint before 30 November 2026. Databricks’ Genie One MCP docs (updated 11 November) deprecate https://<workspace-hostname>/api/2.0/mcp/genie and say it reaches end of life on 30 November 2026. The generally available URL is https://<workspace-hostname>/ai-gateway/mcp-services/system.ai.genie_one_mcp?o=<workspace-id>, and OAuth clients must request the ai-gateway scope instead of genie. n8n’s tile has called the AI Gateway endpoint since 30 October 2026; a credential created before that change keeps a genie token that the new endpoint rejects with “Invalid scope, required scopes: ai-gateway”. An account admin adds the scope; each user reopens the credential, chooses Switch account, and signs in again.
  • Budget agent iterations, not just DBUs. Genie answers asynchronously. The n8n agent calls genie_ask, then genie_poll_response every 2–5 seconds (Databricks’ recommendation) until the turn finishes. n8n’s docs say Databricks states typical turns take 70 to 260 seconds, so one answer can consume 14 to 130 agent iterations. The default Max Iterations of 10 is too low; n8n recommends 30 or more. After a cutoff, Genie can keep running on the Databricks side.
  • Call one known agent by ID with the Databricks node. The MCP tile always lets Genie One choose the agent. To pin a single Genie Agent, n8n points you at the Databricks node instead.
  • Do not pay Genie LLM fees for jobs that should be a workflow. A nightly warehouse refresh, a invoice export or a Slack alert that needs no judgement is usually cheaper as an n8n execution than as an always-on Genie job plus SQL compute.

Before you build, estimate n8n volume with our n8n cost calculator, check finished workflows with the n8n workflow auditor, and compare AI tool prices in our AI SaaS pricing comparison. For other enterprise agent launches, see Microsoft Copilot Studio persistent agents, Salesforce Agentforce Coworker pricing and n8n Nod.

What we couldn’t verify

  • Whether Genie Search is generally available. The 11 November press FAQ lists it as coming soon to preview and beta; the Ontology blog and Search docs call it generally available. We have not tried to enable it in a workspace.
  • A date for always-on Genie and Ontology Manager preview or beta. Databricks only says “coming soon”.
  • DBUs consumed per chat, agent run or always-on job. No official consumption examples were on the pricing or consumption pages we checked, so we have not published a worked monthly bill beyond the DBU arithmetic above.
  • Post-promo Ontology price, and non-US-East DBU dollar rates (the page says they vary by region).
  • Quality and productivity claims (91% Turing benchmark, 1.6×, 10x CEO productivity, HelloFresh’s 75%, the 20,000-organisation and 70% Fortune 500 figures). These are vendor or customer statements.
  • n8n compatibility with always-on Genie jobs (inbox, email alerts, long-running background work). The documented tile covers Genie One question-and-answer over MCP, not the new always-on surfaces.
  • Whether Unity Gateway budgets can cap always-on background compute once that mode ships. Today’s budget docs cover Genie LLM usage only.

Frequently asked questions

What did Databricks announce on 11 November 2026?

The second generation of Genie: an always-on enterprise agent that can run in the background, delegate to sub-agents, and reach people through a priority inbox and email when a decision is needed. The same day Databricks announced general availability of Genie Ontology. The press release is clear that Genie and Genie Ontology are generally available, and that Genie Search, Ontology Manager and the new always-on Genie are coming soon to preview and beta programs. The Ontology blog, published the same day, calls Genie Search generally available. We could not reconcile those two statements from public pages.

How much does Databricks Genie cost?

On the Genie pricing page (AWS, US East, checked 12 November 2026), Genie One, Genie Agents and Genie Ontology are promoted as free for LLM usage until 31 January 2027. From 1 February 2027, named users get 150 free DBUs a month and pay $0.070 per DBU after that, plus underlying compute (for example a SQL warehouse), which is billed separately. 150 DBUs equals $10.50 in US East; the dollar equivalent varies by region. Genie Code already uses the 150 DBU free allowance, with a 25% promotional discount on billed usage until 31 January 2027. Ontology’s post-promo price is listed as coming soon. Service principals do not get the free allowance.

Is always-on Genie generally available today?

Not according to the 11 November press release. That release says Genie and Genie Ontology are generally available to all customers, and that Genie Search, Ontology Manager and the new always-on Genie are coming soon to preview and beta programs. Treat always-on jobs, the priority inbox and email alerts as announced capabilities, not as something every workspace can turn on today.

Does the free Genie promo include SQL warehouse compute?

No. Databricks states that free usage applies to Genie LLM usage only. Compute used to run the work Genie generates, such as a SQL warehouse, is billed separately and is not covered by the 150 DBU allowance or the promo. Unity Gateway budgets track LLM usage, not that compute.

Can n8n talk to Databricks Genie?

Yes. n8n has a Databricks Genie One tile in its MCP servers registry that connects an n8n AI agent to the Genie One MCP server through the workspace AI Gateway endpoint. It uses a Databricks OAuth2 credential with user login (no service principals or personal access tokens on that tile). Databricks is retiring the beta MCP endpoint on 30 November 2026; credentials still using the old genie scope need the ai-gateway scope and a re-login. To call one specific Genie Agent by ID, n8n’s Databricks node is the documented alternative.

What happens to Genie MCP integrations on 30 November 2026?

Databricks’ Genie One MCP docs (updated 11 November 2026) say the beta endpoint https://<workspace-hostname>/api/2.0/mcp/genie reaches end of life on 30 November 2026, after which requests to it fail. Workloads should move to the generally available Unity Gateway URL for system.ai.genie_one_mcp, and OAuth clients should request the ai-gateway scope instead of genie. n8n’s Genie One tile has called the AI Gateway endpoint since 30 October 2026; a pre-change credential keeps a genie token that the new endpoint rejects.

Related reading

Sources

Databricks press release: always-on Genie (11 Nov 2026); Announcing GA of Genie Ontology; Databricks Genie pricing; Monitor Genie cost; Genie budgets; Genie consumption guide; Set up Genie Search; Genie One MCP server docs; Databricks Genie product page; Genie Agents product page; n8n Genie One MCP tile, Databricks credentials, Databricks node and n8n pricing. All checked 12 November 2026 (Asia/Tehran). DBU examples are our arithmetic from the US East list rate.

hamid7777 Avatar

Written & tested by

The Sophrai brief

One calm email when something actually matters.

New tools, real cost breakdowns and the automation news worth your time. No hype, no spam.