▮ExtDB
API & MCP

LiveScope

Real-time AI response reliability scoring. Detects patterns that make AI responses unreliable.

As of June 2026, LiveScope has 7 users in the Productivity category.

EverythingThreadsProductivity
Chrome Web Store ↗.crx
Usersno change0%
7
7
Ratingno change0%
—
— reviews
Reviewsno change0%
—
Version
1.2.32
Manifest V3
90-day change · In the last 90 days this extension 4 version updates, changed permissions.

History

7 snapshots

Tracking since Apr 26, 2026.

7.3252.6799999999999997Apr 26, 2026Jun 8, 2026
View as table
DateUsersRatingReviewsVersion
Apr 26, 2026———1.2.0
May 3, 2026———1.2.0
May 14, 20263——1.2.13
May 20, 20266——1.2.28
May 26, 20265——1.2.30
Jun 3, 20266——1.2.32
Jun 8, 20265——1.2.32
Now7——1.2.32

Changelog

  • May 26, 2026
    description
    LiveScope is an independent AI response reliability scorer. It grades every response you get from ChatGPT, Claude, Gemini, Mistral Le Chat, or Microsoft Copilot — in real time, right next to the chat — and surfaces the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, fabricated specifics, missing context.

You see the score before you act on the response. You learn the cues, not just the verdict.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response score on every reply. Runs on-device by default when you opt in: Chrome's built-in Prompt API (Gemini Nano) first, then a self-hosted WebLLM model (Gemma 2-2B-it via WebGPU, ~1.4 GB cached locally). No response text leaves your browser when local mode is active.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

What's new in v1.2.30:
• Second-judge consensus — when a second model is consulted, its agreement, slight disagreement, or disagreement with the primary verdict now displays directly on the response card.
• ChatGPT selector swap — switched away from the brittle text-message-part hook to the current ChatGPT DOM, restoring reliable scoring after OpenAI's latest UI shuffle.
• Retired the legacy Coach storage keys; existing preferences migrate cleanly.
• Popup type scale and handbook responsive layout polished for small and high-DPI screens.
• Demo-mode hook for in-product onboarding (off by default).



Available in 47 languages with automatic locale detection.

Privacy by design:
• No accounts, no analytics, no tracking pixels, no third-party SDKs.
• When the on-device path is enabled, response text never leaves your browser.
• When disabled, response text is sent to a Cloudflare Worker we operate (blackbird-lens-api.kariem.workers.dev) for scoring only — never retained beyond the request.
• All preferences live in browser storage and never leave your device.

Built by Everything Threads — an independent UK-based AI integrity company. Designed against EU AI Act compliance ahead of the August 2026 obligations. ICO registered: C1896585.

everythingthreads.com
    LiveScope is an independent AI response reliability scorer. It grades every response you get from ChatGPT, Claude, Gemini, Mistral Le Chat, or Microsoft Copilot — in real time, right next to the chat — and surfaces the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, fabricated specifics, missing context.

You see the score before you act on the response. You learn the cues, not just the verdict.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response score on every reply. Runs on-device by default when you opt in: Chrome's built-in Prompt API (Gemini Nano) first, then a self-hosted WebLLM model (Gemma 2-2B-it via WebGPU, ~1.4 GB cached locally). No response text leaves your browser when local mode is active.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

What's new in v1.2.32:
• Settings popup now scales to your display — 320–680px wide based on screen size, instead of a fixed width that felt cramped on big monitors and oversized on small laptops.
• In-page panels (score gauge, Deep Audit card, Coach tile) are draggable again — fixes a regression introduced in early-May builds where positioning rules accidentally blocked the drag behaviour.


Available in 47 languages with automatic locale detection.

Privacy by design:
• No accounts, no analytics, no tracking pixels, no third-party SDKs.
• When the on-device path is enabled, response text never leaves your browser.
• When disabled, response text is sent to a Cloudflare Worker we operate (blackbird-lens-api.kariem.workers.dev) for scoring only — never retained beyond the request.
• All preferences live in browser storage and never leave your device.

Built by Everything Threads — an independent UK-based AI integrity company. Designed against EU AI Act compliance ahead of the August 2026 obligations. ICO registered: C1896585.

everythingthreads.com
  • May 20, 2026
    description
    LiveScope is an independent AI response reliability scorer. It grades every response you get from ChatGPT, Claude, Gemini, Mistral Le Chat, or Microsoft Copilot — in real time, right next to the chat — and surfaces the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, fabricated specifics, missing context.

You see the score before you act on the response. You learn the cues, not just the verdict.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response score on every reply. Runs on-device by default when you opt in: Chrome's built-in Prompt API (Gemini Nano) first, then a self-hosted WebLLM model (Gemma 2-2B-it via WebGPU, ~1.4 GB cached locally). No response text leaves your browser when local mode is active.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

What's new in v1.2.28:
• Gemini selectors fixed — scoring now works on the current Gemini DOM.
• 8-retry cap on empty responses prevents log spam on slow streams.
• Responsive layout — overlay, audit panel, and Coach tile now adapt to laptop, tablet, and split-screen widths.
• Hardened panel positioning so host-page CSS can't crush the overlay to a sliver.
• Popup width fix — no more horizontal scroll or clipped content on Samsung / OS-zoomed displays.
• Switched all injected UI from <body> to <html> to defeat host-page transforms.


Available in 47 languages with automatic locale detection.

Privacy by design:
• No accounts, no analytics, no tracking pixels, no third-party SDKs.
• When the on-device path is enabled, response text never leaves your browser.
• When disabled, response text is sent to a Cloudflare Worker we operate (blackbird-lens-api.kariem.workers.dev) for scoring only — never retained beyond the request.
• All preferences live in browser storage and never leave your device.

Built by Everything Threads — an independent UK-based AI integrity company. Designed against EU AI Act compliance ahead of the August 2026 obligations. ICO registered: C1896585.

everythingthreads.com
    LiveScope is an independent AI response reliability scorer. It grades every response you get from ChatGPT, Claude, Gemini, Mistral Le Chat, or Microsoft Copilot — in real time, right next to the chat — and surfaces the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, fabricated specifics, missing context.

You see the score before you act on the response. You learn the cues, not just the verdict.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response score on every reply. Runs on-device by default when you opt in: Chrome's built-in Prompt API (Gemini Nano) first, then a self-hosted WebLLM model (Gemma 2-2B-it via WebGPU, ~1.4 GB cached locally). No response text leaves your browser when local mode is active.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

What's new in v1.2.30:
• Second-judge consensus — when a second model is consulted, its agreement, slight disagreement, or disagreement with the primary verdict now displays directly on the response card.
• ChatGPT selector swap — switched away from the brittle text-message-part hook to the current ChatGPT DOM, restoring reliable scoring after OpenAI's latest UI shuffle.
• Retired the legacy Coach storage keys; existing preferences migrate cleanly.
• Popup type scale and handbook responsive layout polished for small and high-DPI screens.
• Demo-mode hook for in-product onboarding (off by default).



Available in 47 languages with automatic locale detection.

Privacy by design:
• No accounts, no analytics, no tracking pixels, no third-party SDKs.
• When the on-device path is enabled, response text never leaves your browser.
• When disabled, response text is sent to a Cloudflare Worker we operate (blackbird-lens-api.kariem.workers.dev) for scoring only — never retained beyond the request.
• All preferences live in browser storage and never leave your device.

Built by Everything Threads — an independent UK-based AI integrity company. Designed against EU AI Act compliance ahead of the August 2026 obligations. ICO registered: C1896585.

everythingthreads.com
  • May 14, 2026
    description
    LiveScope is an independent AI response reliability scorer. It grades every response you get from ChatGPT, Claude, Gemini, Mistral Le Chat, or Microsoft Copilot — in real time, right next to the chat — and surfaces the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, fabricated specifics, missing context.

You see the score before you act on the response. You learn the cues, not just the verdict.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response score on every reply. Runs on-device by default when you opt in: Chrome's built-in Prompt API (Gemini Nano) first, then a self-hosted WebLLM model (Gemma 2-2B-it via WebGPU, ~1.4 GB cached locally). No response text leaves your browser when local mode is active.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

What's new in v1.2.13:
• Renamed from Blackbird Scope to LiveScope.
• On-device model upgraded to Gemma 2-2B-it for ~95% strict-JSON reliability (was ~70%).
• New cascade: built-in Prompt API → WebLLM → Worker.
• Sparkline trend graph in the popup tracking your last 50 reliability scores.
• Esc keyboard shortcut to dismiss the in-page overlay.
• Smoother appear/dismiss animations via CSS View Transitions.
• Container queries — overlay adapts to narrow vs wide chat surfaces.
• Per-response feedback now visible on every overlay so you can flag false positives and false negatives directly. False negatives are the most valuable calibration signal.
• Microsoft Copilot and Mistral Le Chat selectors verified live.

Available in 47 languages with automatic locale detection.

Privacy by design:
• No accounts, no analytics, no tracking pixels, no third-party SDKs.
• When the on-device path is enabled, response text never leaves your browser.
• When disabled, response text is sent to a Cloudflare Worker we operate (blackbird-lens-api.kariem.workers.dev) for scoring only — never retained beyond the request.
• All preferences live in browser storage and never leave your device.

Built by Everything Threads — an independent UK-based AI integrity company. Designed against EU AI Act compliance ahead of the August 2026 obligations. ICO registered: C1896585.

everythingthreads.com
    LiveScope is an independent AI response reliability scorer. It grades every response you get from ChatGPT, Claude, Gemini, Mistral Le Chat, or Microsoft Copilot — in real time, right next to the chat — and surfaces the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, fabricated specifics, missing context.

You see the score before you act on the response. You learn the cues, not just the verdict.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response score on every reply. Runs on-device by default when you opt in: Chrome's built-in Prompt API (Gemini Nano) first, then a self-hosted WebLLM model (Gemma 2-2B-it via WebGPU, ~1.4 GB cached locally). No response text leaves your browser when local mode is active.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

What's new in v1.2.28:
• Gemini selectors fixed — scoring now works on the current Gemini DOM.
• 8-retry cap on empty responses prevents log spam on slow streams.
• Responsive layout — overlay, audit panel, and Coach tile now adapt to laptop, tablet, and split-screen widths.
• Hardened panel positioning so host-page CSS can't crush the overlay to a sliver.
• Popup width fix — no more horizontal scroll or clipped content on Samsung / OS-zoomed displays.
• Switched all injected UI from <body> to <html> to defeat host-page transforms.


Available in 47 languages with automatic locale detection.

Privacy by design:
• No accounts, no analytics, no tracking pixels, no third-party SDKs.
• When the on-device path is enabled, response text never leaves your browser.
• When disabled, response text is sent to a Cloudflare Worker we operate (blackbird-lens-api.kariem.workers.dev) for scoring only — never retained beyond the request.
• All preferences live in browser storage and never leave your device.

Built by Everything Threads — an independent UK-based AI integrity company. Designed against EU AI Act compliance ahead of the August 2026 obligations. ICO registered: C1896585.

everythingthreads.com
  • May 3, 2026
    description
    Blackbird Scope is an independent AI behaviour scorer. It grades every response you get from ChatGPT, Claude, Gemini, or Mistral Le Chat — in real time, right next to the chat — on three axes: Faithfulness, Precision, and Reliability (F/P/R).

It flags the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, missing context, incomplete answers. You see the score before you act on the response.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response F/P/R score on every reply.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

Available in 10 languages at launch: English, French, Spanish (Latin America), Portuguese (Brazil), Simplified Chinese, Hindi, Urdu, Russian, Arabic, Bengali. More regional dialects rolling out.

Privacy-first: no personal data stored. GDPR-aligned. ICO registered: C1896585.

Built by EverythingThreads — independent AI behaviour research. everythingthreads.com
    LiveScope is an independent AI response reliability scorer. It grades every response you get from ChatGPT, Claude, Gemini, Mistral Le Chat, or Microsoft Copilot — in real time, right next to the chat — and surfaces the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, fabricated specifics, missing context.

You see the score before you act on the response. You learn the cues, not just the verdict.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response score on every reply. Runs on-device by default when you opt in: Chrome's built-in Prompt API (Gemini Nano) first, then a self-hosted WebLLM model (Gemma 2-2B-it via WebGPU, ~1.4 GB cached locally). No response text leaves your browser when local mode is active.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

What's new in v1.2.13:
• Renamed from Blackbird Scope to LiveScope.
• On-device model upgraded to Gemma 2-2B-it for ~95% strict-JSON reliability (was ~70%).
• New cascade: built-in Prompt API → WebLLM → Worker.
• Sparkline trend graph in the popup tracking your last 50 reliability scores.
• Esc keyboard shortcut to dismiss the in-page overlay.
• Smoother appear/dismiss animations via CSS View Transitions.
• Container queries — overlay adapts to narrow vs wide chat surfaces.
• Per-response feedback now visible on every overlay so you can flag false positives and false negatives directly. False negatives are the most valuable calibration signal.
• Microsoft Copilot and Mistral Le Chat selectors verified live.

Available in 47 languages with automatic locale detection.

Privacy by design:
• No accounts, no analytics, no tracking pixels, no third-party SDKs.
• When the on-device path is enabled, response text never leaves your browser.
• When disabled, response text is sent to a Cloudflare Worker we operate (blackbird-lens-api.kariem.workers.dev) for scoring only — never retained beyond the request.
• All preferences live in browser storage and never leave your device.

Built by Everything Threads — an independent UK-based AI integrity company. Designed against EU AI Act compliance ahead of the August 2026 obligations. ICO registered: C1896585.

everythingthreads.com
  • May 3, 2026
    permissions
    storage, sidePanel
    storage, sidePanel, offscreen
  • May 3, 2026
    name
    Blackbird Scope
    LiveScope
  • May 3, 2026
    host_permissions
    https://chat.openai.com/*, https://chatgpt.com/*, https://claude.ai/*, https://gemini.google.com/*, https://chat.mistral.ai/*, https://everythingthreads.com/*, https://blackbird-lens-api.kariem.workers.dev/*
    https://chat.openai.com/*, https://chatgpt.com/*, https://claude.ai/*, https://gemini.google.com/*, https://chat.mistral.ai/*, https://copilot.microsoft.com/*, https://everythingthreads.com/*, https://blackbird-lens-api.kariem.workers.dev/*, https://huggingface.co/*, https://raw.githubusercontent.com/mlc-ai/*

Permissions & access

Permissions
storagesidePaneloffscreen
Host access
https://chat.openai.com/*, https://chatgpt.com/*, https://claude.ai/*, https://gemini.google.com/*, https://chat.mistral.ai/*, https://copilot.microsoft.com/*, https://everythingthreads.com/*, https://blackbird-lens-api.kariem.workers.dev/*, https://huggingface.co/*, https://raw.githubusercontent.com/mlc-ai/*

Screenshots

LiveScope screenshot 1LiveScope screenshot 2

About

LiveScope is an independent AI response reliability scorer. It grades every response you get from ChatGPT, Claude, Gemini, Mistral Le Chat, or Microsoft Copilot — in real time, right next to the chat — and surfaces the patterns that make AI output unreliable: sycophancy, overconfidence, ungrounded claims, fabricated specifics, missing context.

You see the score before you act on the response. You learn the cues, not just the verdict.

Three tiers, one framework:

• Scope Tier 1 — FREE, unlimited. Instant single-response score on every reply. Runs on-device by default when you opt in: Chrome's built-in Prompt API (Gemini Nano) first, then a self-hosted WebLLM model (Gemma 2-2B-it via WebGPU, ~1.4 GB cached locally). No response text leaves your browser when local mode is active.
• Scope Tier 2 — FREE, 10 session audits per day. Analyses patterns across the whole conversation. Auto-falls back to Tier 1 when the daily limit is hit.
• Scope Tier 3 — PAID or invite. Full audit pipeline for professionals who need defensible, auditable AI use.

What's new in v1.2.32:
• Settings popup now scales to your display — 320–680px wide based on screen size, instead of a fixed width that felt cramped on big monitors and oversized on small laptops.
• In-page panels (score gauge, Deep Audit card, Coach tile) are draggable again — fixes a regression introduced in early-May builds where positioning rules accidentally blocked the drag behaviour.


Available in 47 languages with automatic locale detection.

Privacy by design:
• No accounts, no analytics, no tracking pixels, no third-party SDKs.
• When the on-device path is enabled, response text never leaves your browser.
• When disabled, response text is sent to a Cloudflare Worker we operate (blackbird-lens-api.kariem.workers.dev) for scoring only — never retained beyond the request.
• All preferences live in browser storage and never leave your device.

Built by Everything Threads — an independent UK-based AI integrity company. Designed against EU AI Act compliance ahead of the August 2026 obligations. ICO registered: C1896585.

everythingthreads.com

Technical

Version
1.2.32
Manifest
V3
Size
2.53MiB
Min Chrome
88
Languages
1
Featured
No

Metadata

ID
mkpehkfdfgaaoopmgahmenaagkccnkon
Developer ID
u11029786a8d751b8525abc7d5042aa5d
Developer Email
[email protected]
Created
Apr 25, 2026
Last Updated (Store)
May 23, 2026
Last Scraped
Jun 8, 2026
Website
—

Data sourced from the Chrome Web Store · last verified Jun 8, 2026.