A grounded AI workspace
for MSP operations.
Ask about any client, dashboard, or integration in your tenant. ScopeGrid’s AI reads your live data, knows the page you’re on, renders answers as inline widgets you can pin to dashboards, and keeps every prompt under your provider, your redaction policy, your audit log.
Chatthatdoesthework,wheretheworkhappens.
A side panel that opens beside the page you're already on. Multi-turn, persistent threads, AI-titled, with click-to-pick choice cards.
Italreadyknowswhereyouare.
The workspace knows which dashboard, widget, client, integration, or settings page you're on — and remembers where each thread started.
Groundedinyourtenant.
Tools fetch your real dashboards, widget config, alert history, backup health, security posture. Answers cite which integration the data came from.
Alerts.Reports.Dashboards.
Ask the workspace to set up an alert, a weekly client report, or a brand-new dashboard. Each request returns a previewable, savable spec inline.
Numberscomebackascharts,notparagraphs.
KPI cards, gauges, mini bar / donut charts, ranked lists — inline in the chat bubble. Same widget vocabulary as your dashboards.
Savethemoment.
One click pins the rendered widget onto any of your custom dashboards as a moment-in-time snapshot — perfect for QBR-style tiles.
Tellstheappwheretogo.
"Show me Acme's action items" routes you straight to that client, that tab, with the workspace still open alongside.
Astarterlibrary,filteredtoyou.
New orgs get starter alerts and reports narrowed to integrations they've connected — Veeam failures, secure score drops, weekly recap, QBR.
Pickituptomorrow.
Inline widgets, dashboard previews, and savable spec cards persist alongside the message text. The visuals re-render exactly as before.
Six focused uses of AI you turn on individually.
Each of these is its own feature with its own settings page — enable what's useful, leave the rest off.
The1–3thingsonthisdashboardworthasecondlooktoday.
A dismissible top-of-dashboard banner that highlights anomalies the AI noticed. Threshold-tunable, per-tenant, and never repeats the same callout twice in a row.
PlainEnglishforanywidget,ondemand.
Click the sparkle on any widget and the AI writes a one-paragraph explanation of what it shows and what's standing out — fetched from the actual rendered values, not a generic template.
Lastweek'snoise,triagedintoaparagraph.
One AI-generated summary that turns a wall of alert history into a triage list with concrete next steps. Sent on demand or attached to your weekly recap.
RankedcatalogcandidatesforunmatchedPSAsignals.
Zod-typed structured output, not free text. The AI proposes the top 3 catalog matches with a confidence score; you approve or skip. No hallucinated SKUs.
Aone-paragraphsummaryatthetopofeveryscheduledreport.
Aggressively redacted (names tokenised before they leave your tenant), then drafted into a concise executive summary that sits above the rendered widgets.
Askwhatanintegrationsyncsandgettheactualanswer.
The AI reads our integration catalog — sources, fields, types, capabilities — and answers grounded in real metadata rather than guessing what BackupRadar or CIPP exposes.
Governance
Yours to control
Built for MSPs who need to put AI in front of clients without losing the audit trail.
Bring your own provider
Azure OpenAI (UK South recommended), OpenAI, Anthropic, or any OpenAI-compatible endpoint. Per-tenant key, encrypted at rest.
Per-feature controls
Enable/disable each feature, override the model, set monthly token caps, choose redaction level — all from one settings page.
Per-client opt-outs
Exclude any client from every AI feature. Enforced before data leaves the database for tool-scoped queries.
Default redaction
Email and phone stripped from prompts before they go to the provider. Aggressive mode tokenises contact names → Person_N.
Audit log + usage
Every call logged with hashed prompt, token counts, status, model, cost estimate. Browse by feature, user, status. No raw prompt text retained.
Rate limits
Per-user 30/min and per-tenant 600/min fixed-window limits, counted via the audit log — no extra infrastructure.
How It Works
Roll out in three steps
Connect your provider
Paste an API key (Azure OpenAI, OpenAI, Anthropic, or compatible). Test connection, pick default + fast models.
Enable features per tenant
Turn features on individually with per-feature redaction, monthly cap, and model override. Default is everything off.
Use it in flow
⌘K from anywhere. The palette inherits page context and keeps threads across sessions. Audit log gives you visibility.
Supported Providers
Use the provider your team trusts
Hosted, self-hosted, or anything OpenAI-compatible. Per-tenant configuration, no shared models, no vendor lock-in.
Azure OpenAI
Recommended for UK/EU GDPR-conscious deployments. Use UK South or another approved region.
OpenAI
Direct OpenAI API. Data-processing terms apply per OpenAI's standard DPA.
Anthropic (Claude)
Direct Anthropic API. Strong for long-context reasoning.
Custom (OpenAI-compatible)
Self-hosted Ollama / vLLM, OpenRouter, Together, Groq, LM Studio — any endpoint exposing the OpenAI Chat API.
Logos shown for identification only — ScopeGrid is not endorsed by these providers.
Ready To Try It
Enable AI on your terms
Start with one feature, validate the value, then expand across your tenant. Or hit ⌘K in the demo and just talk to it.