Sevenda is a Chrome Extension (MV3) that records your browser sessions and automatically generates BPMN 2.0 process diagrams and GTM/GA4 analytics insights — powered by Claude Sonnet AI.

It works in two modes: 🧩 BPMN mode maps user flows and processes from real browser interactions; 📊 Insights mode analyzes your tracking setup, identifies missing GTM events, and generates a ready-to-use Tag Plan. No code changes, no SDK, no setup required on the target site.

Sevenda is built for three main audiences:

Developers & Business Analysts — document AS-IS processes, user flows, and API interactions directly from the browser, without manual diagramming.

Marketing & Data teams — audit GTM/GA4 tracking coverage, identify funnel gaps, and generate Tag Plans without manual inspection.

Consultants & Project Managers — produce BPMN diagrams and .docx reports for clients in minutes, and push directly to Jira.

Sevenda is a Chrome Extension Manifest V3 and works on Google Chrome and any Chromium-based browser (Brave, Edge, Arc). It requires access to Chrome DevTools, so it runs on desktop only — not on mobile browsers.

From installation to your first BPMN diagram: under 60 seconds.

The generation itself (Claude Sonnet processing the event stream) typically takes 15–30 seconds depending on session length and API response time. The Analytics Report in Insights mode is instantaneous — it's generated from cached data without an additional API call.

Yes. Sevenda uses <all_urls> host permissions because the target site is not known in advance — you can record any website you navigate to. It does not inject any visible code or modify the target site in any way.

Some sites with very aggressive Content Security Policies (CSP) may limit which events Sevenda can capture — in particular, some XHR/fetch calls may be partially blocked. The BPMN and Insights output will still be generated from the events that were successfully captured.

Yes — Sevenda works on localhost, internal tools, and staging environments just like on public websites. This makes it particularly useful for documenting internal workflows, admin dashboards, and pre-production user flows before they go live.

1. Install the extension from the Chrome Web Store (search for "Sevenda").
2. Click the Sevenda icon in the Chrome toolbar — a popup appears.
3. Go to Settings → AI Models and enter your Anthropic API key.
4. Open Chrome DevTools (F12) on any page — you'll find the Sevenda panel in the tabs.
5. Press Avvia Registrazione and start navigating.

1. Create an account at console.anthropic.com.
2. Go to API Keys and click Create Key.
3. Copy the key (starts with sk-ant-) — you'll only see it once.
4. Paste it in Sevenda → Settings → AI Models → API Key field and save.

Anthropic offers a free credit tier to get started. The key is stored locally in chrome.storage.sync — it never leaves your device unless you explicitly click Generate.

A few things to check:

• Make sure you're on an http:// or https:// page — Sevenda doesn't work on Chrome internal pages (chrome://) or the New Tab page.
• Check that the Sevenda panel is the active tab in DevTools when you press start.
• If the connection indicator shows a warning, try closing and reopening DevTools — the Service Worker keepalive ping reconnects automatically.
• Verify your API key is saved in Settings → AI Models.

Click the Sevenda icon in the Chrome toolbar — the popup shows a toggle between 🧩 BPMN and 📊 Insights. Your selection is saved in chrome.storage.sync and persists across sessions.

Switching modes also changes the event capture filters automatically — Insights mode enables DOM observation and network tracking optimized for tag detection, while BPMN mode prioritizes navigation and interaction events.

Yes. Each team member installs Sevenda on their own Chrome and configures their own API key. Sessions are stored locally per device — there is no shared session storage in the Free and Pro plans.

The Team plan (€29/month) adds a shared session library and team analytics dashboard, allowing up to 10 team members to access and replay each other's sessions.

Yes. Sevenda generates BPMN 2.0 XML that is standards-compliant and compatible with the major BPMN tools:

• Camunda Modeler & Camunda Platform (native .bpmn export with adaptForCamunda())
• Signavio
• Bizagi
• bpmn.io

Sevenda includes automatic XML repair for truncated or malformed output, and validates the structure before rendering with bpmn-js.

Accuracy depends on the quality of the recorded session. A well-structured session — with clear navigation, meaningful interactions, and a short context description — typically produces a BPMN that can be delivered with minimal edits.

Sevenda is designed as a starting point accelerator, not a replacement for expert review. The iterative feedback feature (natural language refinement) is specifically built to let you tune the diagram in seconds — add gateways, rename tasks, restructure flows — without leaving the panel.

From the BPMN panel toolbar you can export:

• .bpmn — BPMN 2.0 XML, ready for Camunda Modeler, Signavio or Bizagi
• .svg — vector image, scalable for presentations and documentation
• .png — raster image for slides, emails, reports
• .docx — full Analisi Tecnico-Funzionale (6-section enterprise report) generated via Claude
• Jira issue — creates a Jira issue with XML + SVG attachments and session metadata
• Camunda .bpmn — adapted export with isExecutable="true" and Camunda namespace

Yes — this is one of Sevenda's core features. The Iterative Feedback section in the panel lets you type natural language instructions to modify the diagram:

"Add an exclusive gateway after the login task with two paths: registered user and guest."
"Rename 'Click Submit' to 'Form Submission — Newsletter'"
"Add an error boundary event on the checkout task"

Sevenda maintains a conversation history of up to 6 turns, so Claude has full context of the previous modifications when applying new ones.

Sevenda captures five categories of events, each color-coded in the live stream:

• NAV — page navigations, History API pushState/replaceState, URL changes
• UI — clicks, form submissions, keyboard interactions, button presses
• NET — fetch and XHR calls (monkey-patched), response status codes and timing
• ERR — JavaScript errors, unhandled promise rejections
• DOM — MutationObserver events: modals appearing/disappearing, dynamic content changes

Each event is scored HIGH/MEDIUM/LOW by relevance and filtered before being sent to the AI.

Yes. In the BPMN panel, click Incolla XML to paste any valid BPMN 2.0 XML and render it in the viewer. You can then use the Iterative Feedback feature to ask Claude to modify it — effectively using Sevenda as an AI-powered BPMN editor even without recording a new session.

Sevenda automatically detects the presence and configuration of:

• Google Tag Manager (GTM ID, dataLayer presence)
• Google Analytics 4 (measurement ID, gtag.js)
• Cookiebot (consent management)
• Hotjar (heatmaps and session recording)
• Other common analytics and tag management tools present in the page scripts

Detection is passive and happens automatically during the session recording — no configuration needed.

The Tag Plan JSON is a structured file containing all the suggested GTM events, triggers, variables, and dataLayer pushes identified during the session analysis. It follows a standardized format that can be:

• Imported directly into Google Tag Manager (via GTM JSON import)
• Shared with a developer as a complete implementation spec
• Included in a client deliverable alongside the Analytics Report .docx

Export it from the Insights panel using the ↓ Tag Plan button.

During the session, Sevenda captures every user interaction (clicks, form fills, navigation, CTA interactions) and compares them against the GTM events actually fired in the dataLayer. Interactions that happened but weren't tracked become tracking gaps.

Each gap is reported with the specific CSS selector and XPath of the untriggered element, so developers can implement the fix precisely. Claude then generates the corresponding dataLayer.push() snippet ready to paste into GTM.

The Analytics Report (.docx) is built directly from the cached Insights data already generated in the current session — it doesn't require a new API call to Claude. It renders instantly from the structured JSON that Claude produced during the Insights analysis.

The BPMN and Insights generation (the initial AI call) take 15–30 seconds because Claude Sonnet is processing the full event stream. Once that's done, all derivative outputs (report, Tag Plan) are immediate.

Yes. From the session replay view (Sessions tab), you can run both BPMN and Insights generation on any saved session — regardless of which mode was active when it was recorded. Switch to the session, then use the BPMN and Insights buttons independently to generate both outputs from the same event data.

All session data is stored locally on your device using the browser's IndexedDB API (database name: flowlens). Your settings and API key are stored in chrome.storage.sync, which is synced across your Chrome devices by Google — not by Sevenda.

Sevenda Lab operates no backend infrastructure. There is no server that receives, stores, or processes your session data. You can delete all sessions at any time from the Sessions tab → Delete, or by uninstalling the extension.

No. Sevenda does not capture:

• Passwords or any type="password" field values
• Credit card numbers or payment data
• Full text content of input fields (only element identifiers: tag, ID, CSS selector, visible label)
• Screenshots or visual recordings of the page
• Personal data beyond what appears in the URL or visible element labels

The captured data is limited to navigation URLs, element identifiers (IDs, CSS selectors), network request URLs with status codes, and DOM structural changes — the minimum required to reconstruct a process flow.

When you click Generate, Sevenda sends the recorded event stream (a structured list of events with timestamps, types, and element identifiers) to the Anthropic Claude API using your own API key.

Anthropic's handling of this data is governed by their Privacy Policy. Sevenda does not see, log, or store the API request or response — the transmission happens directly from your browser to Anthropic's servers.

Yes. Sevenda is designed with a local-first, privacy-by-design architecture:

• No backend infrastructure — zero data transmitted to Sevenda Lab servers
• Explicit user consent required before any AI generation (you must click Generate)
• Full user control over stored data — delete sessions at any time
• No collection of personally identifiable information (PII)

Our full Privacy Policy is available at sevenda.dev/privacy.html. For GDPR-related requests, contact hello@sevenda.dev.

Technically yes — Sevenda works on any website. However, we recommend informing your client that you are conducting a session observation for process documentation purposes, consistent with how you would handle any usability observation or screen-sharing session.

Sevenda does not capture passwords, payment data, or full form content — the risk profile is similar to a consultant observing a screen during an interview. That said, always check your client's internal data policies before recording.

The Free plan includes 10 sessions per month, BPMN 2.0 generation, export in .bpmn and .svg formats, and full access to the Chrome DevTools panel.

Features exclusive to Pro and Team plans: Insights & GTM analysis, .docx report export, .png export, Jira and Camunda integration, and team collaboration tools.

Sevenda uses a BYOK (Bring Your Own Key) model — you pay Anthropic directly for API usage, so costs are transparent and under your control.

Rough estimate: 40 sessions with BPMN or Insights generation costs approximately €15–25 total using Claude Sonnet. The Analytics Report (.docx) is generated without an API call, so it doesn't add to your costs.

You can monitor your exact usage in the Anthropic Console. Sevenda includes adaptive rate-limit management to avoid unexpected throttling.

Yes. There are no long-term commitments. You can cancel your Pro or Team subscription at any time — your access continues until the end of the current billing period, then reverts to the Free plan. No cancellation fees, no data loss.

Yes — we offer a Beta Partnership Program for consulting firms and agencies. Qualified partners receive free access to the Team plan for 60 working days, dedicated onboarding, and direct support from the founder in exchange for structured feedback and a case study.

If you're interested, write to hello@sevenda.dev with a brief description of your firm and use case.