The latest features, improvements, and fixes shipped to Sentrum.
Trial now requires a credit card. Every "Start free trial" CTA lands on /pricing → tier picker → Stripe checkout. Closes the loophole where users could sign up with no CC and get implicit 14-day Pro access. 3-day onboarding grace window so users can connect Meta + complete first sync before paywalls land.
Marketing nav Solutions dropdown now stays open while you move from the trigger to the menu (was closing mid-traverse). Added "Connect to AI" as a top-level nav item — the MCP setup page is no longer orphaned.
Three new blog posts on getting the most out of Sentrum from Claude Code / Cursor / Claude Desktop: a 5-minute tutorial, five prompt patterns paid-social specialists use daily, and a deep-dive on building custom MCP tools on the Developer tier.
Pro, Developer, Agency, and Enterprise workspaces now refresh their Meta data hourly (was daily). Free + Starter keep the daily cadence. The active workspaces you care about most stay current without manual sync clicks.
Replaced the regex-based customer split estimation with Meta-native segmentation. Renders only for Advantage+ Shopping Campaigns (where the breakdown actually exists). New + returning customer share, CPA per segment, ROAS per segment.
New pricing tier for media buyers running their workflow inside an AI agent. 2,000 AI calls / day, custom MCP tool registration, webhook subscriptions, priority email support, Discord community. Everything in Pro plus the developer-focused additions.
Cut initial Meta sync time from 6-8 minutes to ~90 seconds. Date chunks within each entity now run concurrently via durable Inngest step parallelism, plus a higher per-workspace concurrency cap.
Client-ready .xlsx exports from Campaign Deep Dive and the main Dashboard. Branded cover sheet, KPI tiles with period-over-period deltas, daily trends sheet, breakdown sheets with green/red CPR fill. Built on exceljs.
New workspaces now land in the dashboard with an AI-generated diagnosis already waiting. After the initial Meta sync completes, Sentrum picks the highest-spend campaign and runs the multi-pass diagnostic so day-one users see the product working before they click anything.
"Decode your account, layer by layer." New top-level dashboard tab that walks Sentrum AI through the same 7-section diagnostic we use internally: Spend Health → KPI Volume → Efficiency → Supporting Metrics → Mix Shifts → Recent Edits → Recommended Actions. Single LLM call, structured output, ~20 seconds.
No more silent charts. The Attribution tab now opens with a "Sentrum AI · Diagnose this attribution" CTA that returns a paragraph-level diagnosis covering over-credit risk, retargeting attribution inflation, and diminishing-returns campaigns. Same numbers as the chart cards below — synthesized.
New "Sentrum AI · {verb}" CTA pattern with a Sparkles glyph and purple-to-blue gradient. Adopted on Campaign Deep Dive empty/loaded states, Attribution AI, Blueprint, Anomaly Dashboard, Creative Library AI Analysis, Creative Comparison. The chrome reads consistently anywhere the AI is talking.
AI insights now use current Meta terminology: Andromeda (not "the algorithm"), Advantage+ Shopping (not "smart shopping"), ODAX objectives, AEM, CAPI, EMQ, Performance 5. Diagnoses also handle ASC-specific dynamics — audience break-outs are limited, frequency caps less direct, etc.
Refreshed every model string across the AI stack to the current generation. Sonnet 4.5 for narrative passes, Gemini 2.5 Flash for creative vision, GPT-5 / GPT-5-mini for analytical passes, Whisper Large v3 for transcripts. ~+15% per-call cost for materially better quality.
Hidden surfaces that depend on data sources we don't have at launch (Shopify-revenue features — Blended MER, Revenue Gap, regex Customer Splits — plus the Cannibalization Detector and Watchlist). Kept the Atria-creative suite (Boards, Saved Ads, Hook Library, Brief Builder, Concept Report) intact.
The Rolling Reach card was summing daily reach, which double-counts the same users across days. Now uses Meta's period-unique reach + fresh-reach ratio per campaign so the number actually means what it says.
/settings/usage was reporting $0 of AI cost because every call site was passing inputTokens: 0 to the usage tracker. Now every Anthropic, OpenAI, and Gemini SDK call threads real token counts through to the cost ledger. Dashboard reflects actual LLM spend.
New reference page at /docs/mcp/prompting-guide. Tool reference for all five MCP tools, prompt patterns (Diagnose / Compare / Hypothesize / Narrate), daily/weekly/creative workflows, troubleshooting. The companion piece to /connect-to-claude.
Complete visual refresh of every public marketing surface — landing, pricing, blog, comparison pages, /connect-to-claude. New design tokens (Plus Jakarta Sans display, DM Sans body, Geist Mono labels), dark-first palette with light parity, gradient hero treatments. Same product, sharper presentation.
Sentrum now exposes its diagnostic engine as an MCP server at sentrum.app/api/mcp. Five tools: getCampaignDiagnostics, getCreativeAnalysis, getAccountContext, getRecommendations, askQuestion. Any MCP-capable AI client (Claude Code, Claude Desktop, Cursor, ChatGPT) can read your Meta ad account through the same diagnostic layer that powers the dashboard.
Replaced the single-call insight generator with a 4-pass pipeline: anomaly detection → root cause diagnosis → recommendations → narrative synthesis. Each pass receives only the context it needs, producing materially deeper analysis than the monolithic approach. Powers Daily Pulse, Campaign Deep Dive, Blueprint, and Attribution AI.
Completely rebuilt creative analysis with AI-powered hook rate scoring, hold rate tracking, fatigue detection curves, and concept clustering. Understand which creative elements actually drive performance.
Go beyond ROAS with true profit visibility. Set product margins, configure attribution models, and split metrics between new vs. returning customers. See real profit per campaign, not just revenue.
A unified, real-time feed of AI-detected insights, anomalies, and recommendations across your entire account. No more digging through tabs — the most important signals surface automatically.
Invite team members with granular role-based permissions. Admins, editors, and viewers — control who can see data, modify settings, and manage integrations across your workspace.
Push alerts and anomaly notifications to Slack channels or custom webhook endpoints. Configure per-alert routing, severity thresholds, and notification schedules.
Press ⌘K to instantly search campaigns, navigate pages, trigger exports, or run actions. Power-user navigation for the keyboard-first workflow.
Statistical anomaly detection across all key metrics — CPM spikes, CPA drift, CTR drops, and spend anomalies. Get alerted before small issues become expensive problems.
AI-powered budget optimization recommendations with real-time pacing tracking. See which campaigns are overspending or underspending, with suggested reallocations.
Export any data table — campaigns, ad sets, ads, or creatives — as a CSV file with all metrics properly formatted. Available via the UI and a new API endpoint for automation.
Generate presentation-ready post-campaign recaps with one click. Includes top-line performance vs benchmarks, audience learnings, creative deep dives, and actionable next steps. Available at the campaign, ad set, and ad level.
Sentrum's Ask AI now explains the "why" behind every performance shift — grounded in your actual account data. Ask questions like "why did CPA spike last week?" and get answers referencing auction dynamics, creative fatigue, placement mix shifts, and learning phase effects.
The core of Sentrum is live. Real-time flags surface CPM spikes, CPA drift, CTR drops, creative fatigue signals, and budget inefficiencies as they happen — not after you manually dig through Ads Manager. Flags include severity levels, root cause context, and recommended actions.
Get notified when we ship new features and updates.