Chrome Extension // Manifest V3 v4.0

Automate the browser
with plain English.

Tell Sentinel Override what you need. It plans the steps, sees the page, and executes clicks, typing, and navigation — autonomously.

sentinel-override — agent session
> Log into my GitHub and star the top 5 repos in my feed
[01] Navigating to github.com/login
[02] Found username field — typing credentials
[03] Submitting login form
[04] MFA detected — paused for authorization
[05] Auth complete — navigating to feed
[06] Staring repo 1/5: langchain
[07] Staring repo 2/5: rust
[08] Staring repo 3/5: deno
[09] Staring repo 4/5: ollama
[10] Staring repo 5/5: zig
Task complete. 5 repos starred. Audit log saved.

Three commands. Zero scripts.

No Selenium. No Playwright. No Puppeteer. Describe the outcome and the agent handles the execution.

STEP 01

Describe your goal

Open the command palette with Cmd+K and type what you need in plain English. The agent receives your intent.

STEP 02

Agent plans the path

Sentinel scans the page using its vision system with Set-of-Mark — every interactive element is identified, labeled, and evaluated against your goal.

STEP 03

Executes autonomously

Clicks, typing, navigation — every action is executed with optional approval gates. Background tabs stay non-interrupting. Resume from any checkpoint.

Built for control. Engineered for trust.

Every feature is designed around observability, safety, and operational precision.

Vision-first with Set-of-Mark

Every element on the page is labeled and mapped. The agent sees what you see — buttons, forms, links, modals — and interacts with them by reference, not brittle selectors.

Command Palette

Activate with Cmd+K from any page. Type your intent, review the plan, and let the agent execute. Keyboard-first, no mouse required.

Approval Gate

Every action can require explicit review before execution. Approve, modify, or skip individual steps. Full control over what the agent does on your behalf.

Forensic Audit Log

Every action is timestamped and logged with full context. Export to JSON or CSV for compliance review, debugging, or forensic analysis.

JSON / CSV EXPORT

Checkpoint & Resume

Tasks save progress at every step. If something interrupts the session — a crash, a close, a network drop — resume exactly where you left off.

MFA Auto-Detection

When the agent encounters a multi-factor authentication prompt, it pauses and waits for your manual input. Never gets locked out. Never bypasses security.

Source-Cited Outputs

Every piece of data the agent extracts is cited back to its origin — exact URL, element, and timestamp. Full provenance chain for every output.

Tenant Lockdown

Restrict agent operations to approved domains only. Enterprise-ready isolation that prevents the extension from acting outside your configured boundaries.

ENTERPRISE

Background Execution

Runs in background tabs without stealing focus. Keep working while the agent handles tasks in parallel. Non-interrupting by design.

14 built-in themes. Zero configuration.

From Matrix green to Tron blue. Match your workflow aesthetic — or build your own.

Matrix · Tron · Cyberpunk · Vaporwave · Synthwave · Neon · Emerald · Crimson · Cobalt · Terminal · Phosphor · Ember · Monochrome · Override

Your model. Your keys. Your rules.

Bring your own API key from any of 20+ supported providers. Sentinel Override never proxies or stores your credentials.

OpenAI
Anthropic
Google Gemini
xAI (Grok)
DeepSeek
Mistral
Meta Llama
Cohere
Perplexity
Together AI
Fireworks
Ollama (local)
OpenAI-compatible endpoint support · Local models via Ollama · BYOK architecture

Ready to override your workflow?

Available now on the Chrome Web Store. One install. One permission. Full browser automation from day one.

Manifest V3 compliant
No data collection
BYOK only
Open-source core