OpSage is a platform in active development, shaped by the realities of running restaurants at scale.
These release notes capture the full history of what ships — new features, architectural improvements, bug fixes, and security updates — so operators, IT teams, and anyone evaluating the platform can see exactly how it has evolved and where it is headed.
Release 0.9.9 — OpSage Meets Your Team in Slack
Released May 26, 2026
OpSage 0.9.9 brings the platform directly into Slack, letting store managers, district managers, and ops teams ask questions and receive the same role-scoped, restaurant-aware answers they get in the web app — without leaving the tool they already work in. This release also completes the migration of all customer-facing surfaces to opsage.com and delivers a measurable speed improvement to AI Chat across both web and Slack.
Technical Detail
The OpSage Slack app uses a one-time admin install paired with a per-user /opsage connect command that links each teammate's Slack account to their OpSage identity via the same enterprise SSO already in use. No separate passwords, no shared logins. Once connected, users send OpSage a direct message and receive answers through the same reasoning pipeline, semantic layer, and role-based access controls as AI Chat on the web. A district manager sees their district; a store manager sees their store — regardless of which device or surface they are asking from.
Under the hood, AI Chat now runs on the same tool surface that powers OpSage's Intelligence Layer connections to Claude, ChatGPT, Gemini, and Slack — one shared brain, one shared security envelope. This architectural consolidation is also the source of the speed improvement: every turn in AI Chat, on both web and Slack, is now noticeably faster as a structural result of the unified surface.
The opsage.com migration moves the app to app.opsage.com and shifts invitations, password resets, alerts, and Daily Briefings to @opsage.com sender addresses. Toast and other POS callback URLs have been updated. Previous convx.com addresses remain live for 30 days to cover bookmarks, saved emails, and existing POS connections. Two report fixes ship in this release: actual metric values now render correctly alongside targets and variance in Business Reports, and percentage-typed custom variables in report templates now compute and display correctly. The Slack app is included in all OpSage subscriptions with no per-user fees.
Blog:
Your Team Is Already in Slack. Now OpSage Is Too.
Released May 18, 2026
OpSage 0.9.8 makes Business Reports fully customizable. Operators can define their own calculated metrics, rename fields to match internal vocabulary, and annotate templates with per-field descriptions. PDFs now render in brand colors with the operator's logo, operational observations travel with exports, and a guided onboarding flow goes live for new workspaces.
Technical Detail
The custom metrics feature adds a formula editor to the report template designer. Operators reference existing report fields by name and build arithmetic expressions, ratios, and percentages. Variance against target works identically to built-in metrics. Custom metrics flow into PDF and CSV exports automatically, and when OpSage generates an AI analysis narrative, it includes commentary on custom metrics alongside the built-in ones — eliminating the need to export to Excel to add calculations after the fact.
Field labels in report templates are now fully editable. Renamed labels propagate to the report viewer, PDFs, CSVs, and AI analysis narrative. Formulas in custom metrics automatically reference fields by their new names. Historical reports continue to render correctly — the rename changes labels only, not underlying data. Per-field descriptions can also be added to any template; they appear as tooltips in the report viewer and as captions in PDFs, giving report readers context on what each metric means without a separate reference document.
Branded PDF reports pull from the workspace Brand Kit: logo in the header, brand colors on section headers and variance highlights, workspace name in the footer. When a report is exported to PDF or Excel, any operational observations attached to that report now come along, grouped by domain and timestamped exactly as they appear in the in-app viewer. New workspaces signing up after this release go through a context-aware guided onboarding flow that skips fields already filled in. Restaurants can also be added directly from the master-data locations page post-onboarding without returning to the flow. All features in this release are included in the OpSage subscription with no per-user fees.
Released May 13, 2026
OpSage 0.9.7 solves one of multi-unit operations' most persistent reporting problems: the same burger with three names across three locations, the same loyalty customer appearing as five records, the same discount labeled differently per terminal. The platform now reconciles menus, customers, and discounts into consistent reporting categories automatically — without modifying POS records — so weekly reports and AI analysis return comparable numbers across the entire footprint.
Technical Detail
Menu normalization groups items into consistent reporting categories across all locations and brands. AI proposes the initial grouping; operators accept or override. The reconciliation happens on the reporting side only — source POS records are not modified. Customer deduplication links guest records so a single customer rolls up consistently across every location where they appear, eliminating the inflated unique guest counts that result from POS-level record fragmentation. Discount normalization categorizes discounts consistently as line discounts, ticket discounts, or promotional comps, so variance analysis and AI questions about discount impact return comparable numbers regardless of how the discount was entered at the terminal.
This release also includes seven bug fixes: AI Chat now reliably returns observations on date-range queries; date extraction from natural-language observation text resolves correctly against the location's local timezone; the date range filter on dashboards and reports persists across page refresh; voice observation capture handles a wider range of audio inputs; the settings sub-tab is preserved across navigation and refresh; accessibility improvements were made in User Profile; and restaurant creation no longer fails on certain brand and region combinations. Four proactive security controls ship in this release: enhanced user-level identity permissions at the data layer, automatic log retention and purge with sensitive information stripped before write, service-to-service access via role-based credentials rather than fixed keys, and heavy reporting jobs running in dedicated workspaces isolated from the application and AI-client connections.
Blog:
One Name for the Same Thing, Across Every Location
Released Spring 2026
OpSage 0.9.6 launches the OpSage Intelligence Layer in production — a standards-compliant MCP server that connects Claude, ChatGPT, and Gemini directly to your operational data. Operators do not replace their AI tools; they make them restaurant-aware. The same data, semantic layer, and role-based access controls that power AI Chat inside the app are now available inside whichever AI client your team already uses, authenticated via enterprise SSO and governed by the same per-user permissions.
Technical Detail
Setup is a single configuration step: generate a token from your OpSage account, paste the configuration snippet into your AI client, and the connection is live. Sign-in flows through the operator's existing enterprise SSO so every team member uses their existing credentials. Tokens are short-lived and tied to the requesting user's identity; first-time users get a stub workspace created automatically. Twenty-plus question types are available at launch, covering dashboards, revenue trends, reviews and sentiment themes, alerts, integration and job status, and configuration items.
Users can also capture operational observations directly from their AI client. OpSage stores the observation identically to one entered in AI Chat inside the app — same classification pipeline, same structured storage, same verbatim text preservation. Beyond reads, AI clients can handle low-risk operational housekeeping: adding a restaurant, adjusting a business profile, refreshing day parts, sending a team invitation. Sensitive operational data changes remain inside the OpSage application.
Reports gain performance targets in this release. Set targets at location, region, or brand level; OpSage calculates variance automatically; update a target and the new value carries forward to every future report. The observation form now pre-populates brand, region, and restaurant context from the report currently in view. Security controls include SOC 2 Type 2-certified identity provider, short-lived tokens, per-request user identity enforcement, and per-token rate limits on AI-client connections. The Intelligence Layer connector, AI Chat, observations, reports, and scheduled distribution all carry no per-user fees.
Blog:
You Already Have an AI. Now Make It Actually Know Your Restaurants.
Released Early 2026
OpSage 0.9.5 introduces Operational Observations — a structured way to capture what managers already know but have had no place to record. Managers describe events in plain language via AI Chat, typed or spoken, and OpSage classifies, structures, and stores the observation so it flows into weekly reports and anomaly explanations automatically. Voice input, alert-linked observations, and the ability to correct AI classifications put the manager in control of the context behind the data.
Technical Detail
When a manager enters an observation — "Tomato prices jumped to $95 a case on Tuesday's order," "Threw out four burger patties, shelf life," "Dishwasher broke at 6pm, back up by 9" — OpSage classifies by domain (food cost, labor, equipment, waste, staffing, reviews, customer), extracts entities (vendors, items, prices, employees, locations), and stores structured records. The original text is always preserved verbatim; AI classification is additive and never overwrites the manager's own words.
Every observation shows the AI's domain tag in chat. If the classification is wrong, the manager can tap the tag to select the correct domain or simply tell OpSage in natural language — "actually that was a labor issue, not food cost" — and the classification updates immediately. When an anomaly alert fires, users can reply directly in AI Chat; the explanation is captured as an observation linked to that alert, so the reason lives with the data rather than in an email thread or group chat.
Voice input works throughout AI Chat — line walks, stockroom, the car after a site visit. Observations surface in weekly Business Reports grouped by domain, so operational context travels with the numbers. Three report crash fixes ship in this release, including one that had blocked custom templates from opening entirely. Security additions include input sanitization against SQL injection in search and observation text, and role-based access control for the Reports module. All observations, voice input, scheduled distribution, and the reports framework are included in the OpSage subscription with no per-user fees.
Blog:
Capture the Why Behind the Numbers
Released Early 2026
OpSage 0.9.4 introduces Business Reports — auto-populated, structured weekly performance reviews that replace the spreadsheets managers compile by hand. Sales, labor, review, and cost data pull directly from connected systems, display actual vs. target variance in color-coded format, and include AI-generated analysis explaining what moved the numbers. Reports export to PDF or CSV and deliver automatically via scheduled email, with no per-user fees for access or distribution.
Technical Detail
Business Reports pull metrics automatically from every system already connected to OpSage. Every metric displays actual against target with color-coded variance highlights; targets are editable and recalculate variance immediately when changed. Manual data entry is supported uniformly for metrics where automated data is not yet available. Reports work at any scope — location, brand, region, or all locations — and any reporting period. Targets carry forward from prior periods when not explicitly set for the current period.
AI analysis is generated on demand for any report. OpSage identifies highlights and concerns in plain language, drawing from the full cross-domain data set available to the platform. Analysis can be regenerated any time after entering new data or changing targets. PDF export includes a branded header with report name, period, and scope; all metrics, targets, variance, AI analysis, and annotations. CSV export is Excel-ready. Scheduled email delivery supports internal team members and external recipients — auditors, franchisors, investors — at no additional cost. There are no per-user fees for report access or distribution.
This release also requires at least one region when creating a brand, ensuring the full brand-region-location hierarchy is established from the start. Several bug fixes ship: AI Chat no longer carries metric data from a previous query into a new one; dashboard metrics refresh correctly when changing brand or location; adding a metric no longer creates a duplicate tile; comparison chart alignment and overlap at narrow widths are resolved; and an error when removing a metric from the dashboard is fixed.
Blog:
The Weekly Report That Builds Itself
Released Early 2026
OpSage 0.9.3 adds the full cost picture to the dashboard — including Total Prime Cost, the combined labor and food cost metric that drives restaurant profitability — alongside daily and total inventory cost visibility. Custom date range selection lets operators analyze any promotional period, holiday week, or arbitrary timeframe. AI Chart generation becomes more robust with additional chart types, intelligent labels, and consistent data-to-chart fidelity. A new Labor Percent of Sales anomaly alert type rounds out the release.
Technical Detail
Three new dashboard metrics are available: Total Prime Cost (labor plus food cost combined), Total Inventory Cost (current inventory value), and Daily Inventory Cost (day-over-day inventory spend). These sit alongside existing sales, labor, and review data, completing the operational cost picture in a single view. Custom date range selection is now available on the dashboard alongside expanded one-click preset periods, enabling analysis of promotional periods, holiday weeks, or any timeframe relevant to the operation.
Chart generation in AI Chat adds additional chart types and intelligent labels. Charts render consistently and data displayed in charts matches the accompanying text response exactly. The new Labor Percent of Sales anomaly alert type automatically detects when the labor cost ratio deviates from historical baselines and delivers the alert via email and SMS, configurable from alert preferences. Two bug fixes ship: improved error messages when API requests fail or are blocked, and resolved access issues for users invited to existing accounts.
Blog:
Prime Cost Visibility and Flexible Dashboards
Released Early 2026
OpSage 0.9.2 introduces AI-powered competitive analysis and business profiles. Operators define their brand's identity — cuisine type, price tier, service style, hours — and OpSage uses that profile as the foundation for automatically discovering competitors, generating competitor profiles, and enabling side-by-side comparisons. Business profiles and the competitive landscape are immediately accessible in AI Chat. Read-only inventory and menu data views round out the release.
Technical Detail
Business profiles capture brand identity directly in OpSage — cuisine type, price tier, service style, hours, and additional operational attributes. For multi-brand operators, each brand gets its own isolated profile, keeping data and analysis cleanly separated. The profile becomes the foundation for competitive analysis, AI-powered insights, and future benchmarking features.
Competitive analysis runs automatically: OpSage discovers competitors near each location, generates profiles covering operations, positioning, and reviews, and enables side-by-side comparison to surface where the operator leads and where opportunity exists. Competitors can also be added manually. Business profiles and the competitive landscape are available in AI Chat immediately — operators can ask about competitive positioning, price tier comparisons, and cuisine type mix for any market, with the AI combining operational data and business profile for full-context answers.
The Master Data section adds read-only views of inventory and menu data synced from integrations: ingredients with categories and tags, and menu items with recipes, ingredients, and measurements. Six bug fixes ship in this release, addressing the Performance Targets page load failure, integration sync not starting after onboarding, competitor discovery generating fewer profiles than selected, the location selector not showing city names in daily reports, daily report generation failures under high API load, and a data pipeline issue affecting AI Chat and Sentiment Analysis.
Released Early 2026
Release 0.9 is a significant upgrade to the core AI experience. A new multi-step reasoning engine replaces the prior query architecture, enabling complex multi-part questions, true conversational context across follow-up questions, and industry research alongside internal data. Conversation threads let operators organize analytical work by topic. Labor analytics land in AI Chat with support for 15-plus question types. Integration management gains per-channel granularity with Toast POS as the reference four-channel model.
Technical Detail
The new reasoning engine, built on LangGraph, handles complex multi-part questions in a single message and returns a complete answer for each part. Conversational context now persists across follow-up questions — "Show me sales for all locations" followed by "Do the same for Downtown" is understood correctly. When a question goes beyond internal data, the AI researches external sources and includes that context alongside operational data. Ambiguous questions trigger a clarifying question before the AI runs a query rather than returning potentially inaccurate results.
Conversation threads support named, renamable, and deletable threads organized by topic. A single question can generate multiple charts side by side for visual comparison. Labor analytics in AI Chat support labor cost as a percentage of sales by period and location, which locations exceed labor targets, Sales Per Labor Hour (SPLH) by restaurant, regular vs. overtime hours comparisons across locations, overtime trends month over month, and weekday vs. weekend labor efficiency.
Integration management now surfaces per-channel visibility and control. Toast POS is presented as four distinct channels — Orders, Menu, Labor, and Dining Options — each with independent sync status, schedule, and credentials. A redesigned unified Manage Integrations experience consolidates credential configuration, data types, and restaurant mapping. Two bug fixes ship: integration job statuses now load correctly, and the integration progress bar no longer stalls at 0% during active syncs.
Released February 2026
OpSage 0.8.5 brings labor data fully into the platform for the first time. Employee-level records, standardized job roles, jurisdiction-based overtime calculation, and historical labor performance trends are now live — enabling Labor % of Sales, Overtime %, and Total Labor Cost as real-time dashboard metrics. Labor targets can be set at the company level with per-location adjustment. Anomaly detection now covers labor performance alongside the existing revenue and transaction metrics.
Technical Detail
Labor data integration connects employee-level records, maps job roles, and applies overtime calculation based on jurisdiction rules — including daily and weekly thresholds where required. Historical labor performance trends are available for reporting and analysis immediately. Standardized employee and job records ensure a single employee record is used for reporting across locations and that job roles are consistent for cross-location comparisons. Labor % of Sales, Overtime %, and Total Labor Cost are now visible on dashboards as at-a-glance metrics.
Labor cost targets can be configured at the company level with the ability to adjust by location. Targets establish expected performance ranges, power performance monitoring, and support proactive labor management. Labor performance is now included in anomaly detection — alerts trigger when labor performance deviates meaningfully from expected ranges, enabling earlier identification and response. Two fixes ship: a Team tab crash is resolved, and the Clear Cache button has been removed from the UI.
Released February 24, 2026
Release 0.8 introduces a full suite of automated anomaly detection — seven alert types covering revenue, transaction count, average check, menu mix, zero sales, and product velocity — along with the first Brand Kit capabilities for white-label deployments. Advanced Daily Reports gain AI-generated summaries and baseline comparison data. AI Chat improves across empty result handling, multi-turn context retention, natural language matching, and multi-language support.
Technical Detail
Seven anomaly alert types launch in this release: Daily Revenue Change, Daypart Revenue Deviation, Transaction Count Anomalies, Average Check Size Deviation, Menu Item Mix Distribution Shift, Zero Sales Detection, and Product Velocity Change. Anomalies are surfaced with tooltip context and are role-based — users see only alerts relevant to their access level. Alerts deliver via SMS and the Reports modal. Users configure notification preferences directly in the app, selecting which alert types trigger and how they are delivered.
Brand Kit capabilities enable restaurant groups to deploy OpSage under their own brand with a custom logo and custom platform name replacing "OpSage" across the interface — the foundation for broader white-label capabilities in subsequent releases.
Advanced Daily Reports add AI-generated summaries and analysis to the report modal, baseline comparison reports for net sales and daily totals, average order metrics, and an updated tornado chart with AI-powered insights. AI Chat improvements include better handling of empty query results with meaningful guidance when no data matches, improved context retention across multi-turn conversations, case-insensitive natural language query matching, and enhanced multi-language support with automatic language detection. Bug fixes address admin permission issues for integrations and reviews, negative value display in pie charts, zero-result query display, and SQL generation fallback handling.
Download, email, print and share. Everyone on your team will see exactly why you're so excited about OpSage™.
