AI Puffer – AI Chatbot & MCP Server for Claude and ChatGPT
AI Puffer (formerly AI Power) lets ChatGPT and Claude work on your WordPress site through MCP, using your chosen safety level. It also brings an AI chatbot, content writer, automations, AI forms and image generation to your dashboard. Start with 25 free AI credits monthly. Connect AI Puffer Cloud with no API key. Free credits are available to eligible accounts with a verified email. You can also use your own OpenAI, Anthropic Claude, Google Gemini, xAI Grok, DeepSeek, OpenRouter or Azure OpenAI key. Documentation · How Cloud works · Pricing · User reviews Manage your site from ChatGPT or Claude (MCP) Open Settings → Connections → Manage my website to connect ChatGPT, Claude, Cursor, Claude Code or another MCP client. Assistants can read and search content, create drafts and publish posts. Standard asks for WordPress approval before publishing or supported plugin changes. Trusted is the default for new connections and runs supported changes directly, including publication, plugin changes and deletion. Existing Trusted connections keep their previous approvals until you select Trusted again in Settings. The available actions depend on the site’s installed plugins and the permissions you grant. Access follows the connecting WordPress user’s permissions and can be paused or disconnected. Website actions are counted separately from AI credits; your assistant’s own subscription or model charges are separate. This feature is off until you explicitly connect it. Supported WP-CLI-compatible commands let assistants inspect the site and manage posts, images, categories, tags, comments, classic navigation menus, settings, users, roles, cache and scheduled events without SSH or a WP-CLI installation. Official plugin/theme/core administration and bounded database inspection/search-replace commands are included. Ask your assistant for wp help to see the supported subset and flags. Commands use the same permissions, approval settings and website-action allowance as other website tools; shell scripts and full WP-CLI compatibility are not included. Cloud or your own API keys AI Puffer Cloud: connect in the setup wizard or Settings → AI. Supported models cover text, image generation and analysis, speech, transcription and embeddings. Credit use varies by model and request size, with a minimum per text request. One credit is not always one message, article or image. Image editing, video, realtime voice and web search currently require your own supported provider key. Your own API keys: your site sends requests to the provider you choose, which bills your provider account. OpenAI-compatible endpoints are supported too. Use Cloud and your own keys side by side, with a model picker for each chatbot, form, task or tool. Included in the free plugin This edition includes the following core modules without a paid license. AI requests use your own provider account or AI Puffer Cloud credits. Website management is optional and includes 2,000 monthly website actions on the Free allowance, shared across your connected sites and assistants. AI chatbot that answers from your content Create unlimited chatbots with their own models, instructions, colors, welcome messages and conversation starters. Use posts, pages, products and FAQs as knowledge. Show a bot as a popup, site-wide, or with a block or shortcode. Voice input, read-aloud replies and image input work with supported models. Web search and Google Search grounding require your own supported provider key. AI content writer and Content Assistant Draft posts from topics and keywords, one at a time or from a batch or CSV list. Use reusable prompt templates, choose draft or publish status, and add featured or inline images from AI providers or stock photos. Start from Posts → Generate New Post. Write Gutenberg-ready content, with meta descriptions and focus keywords where supported by Yoast SEO, Rank Math, All in One SEO or The SEO Framework. The Content Assistant improves titles, excerpts, content and tags in the block editor, classic editor and post lists. Automations: scheduled content and auto blogging Create recurring or one-time automation tasks from topics or CSV, with draft, schedule or publish options. Reply to comments with replies held for approval by default, or add new posts to your knowledge base using category filters. Review item status in the queue. Server cron is supported when WP-Cron is disabled. AI forms Build forms with a drag-and-drop editor, templates and prompts that use visitors’ field values. Choose a model and knowledge base for each form, and enable web search with your own supported key. Embed with a block or shortcode, and import or export forms. AI image generator Enable Images in Settings → Modules. Generate images with OpenAI GPT Image, Gemini image models, Grok Imagine, Replicate, OpenRouter, Azure or Cloud. Image editing works with supported providers and models. Google Veo video needs your own Google key. Search stock photos from Pexels and Pixabay, or embed a visitor image generator with a block or shortcode. These tools are included in the free plugin; provider accounts or Cloud credits cover AI usage. Knowledge base and semantic search AI Puffer splits content into passages and retrieves relevant passages for a question. It does not train a new AI model. Index posts, pages, products, custom post types, Q&A pairs and pasted text, including page-builder content. Use “Add to knowledge base” from post lists or embed semantic search on your site. Choose Local storage in your WordPress database, OpenAI vector stores, Google File Search, Pinecone, Qdrant or Chroma. External stores and embeddings require compatible provider connections. WooCommerce product tools and visitor credits Improve product titles, descriptions, short descriptions and tags with the Content Assistant; answer questions from product pages; and sell visitor credit packages for your chatbots, forms and image generator. Visitor credits are balances managed on your site, separate from AI Puffer Cloud credits. WooCommerce is optional. Usage, permissions and integrations Set guest, user and role limits with daily, weekly, monthly or no reset. Review conversation and usage logs, credit balances and access through the Role Manager. Word and IP blocklists, IP anonymization and optional moderation help manage requests. Use the REST API, signed event webhooks, three blocks and five shortcodes. WordPress AI Connectors are opt-in and let compatible built-in AI features use your configured providers. Optional upgrades The free plugin works on its own. Paid subscriptions add: File sources and uploads for knowledge bases, chatbots and forms. RSS, URL and Google Sheets content sources, scheduled post improvements and Smart SEO tools. Realtime voice, advanced chatbot triggers, external chatbot embeds, multi-step forms and PDF exports. Additional app integrations, Ollama, automatic log retention and higher monthly Cloud allowances. See the feature comparison and current plans for the full list. All subscription tiers unlock the same additional plugin features; their monthly allowances differ. Supported providers AI Puffer Cloud: available models from OpenAI, Anthropic, Google and others, using AI credits. OpenAI: GPT text models, GPT Image, speech, transcription and embeddings. Anthropic: Claude Sonnet, Opus and Haiku. Google: Gemini text and image models, Veo, speech and embeddings. xAI: Grok text and image models. DeepSeek, OpenRouter (including Mistral models), Azure OpenAI and OpenAI-compatible endpoints. Media services: ElevenLabs voices, Replicate images, and Pexels and Pixabay stock photos. Models and capabilities depend on the selected provider and its current catalog. Privacy and data flow AI requests are sent when you use a connected provider. Visitor messages go to the provider selected for that chatbot. Chat logs, usage records and Local knowledge are stored in your WordPress database. Text you index is processed by the chosen embedding provider and, with an external store, stored there. The External services section below explains connection, licensing and diagnostic requests and links each service’s terms and privacy policy. External services The plugin contacts external services when you connect or use them, manage licensing or checkout, or submit optional deactivation feedback. Cloud setup and connection diagnostics are sent after you accept the Cloud connection terms and click Connect. An AI request contains what the feature needs: the prompt and instructions, chat or form input with recent conversation history, the content being written or improved, text you add to a knowledge base, images or audio you submit, and model settings. Each API key goes only to its own provider. AI Puffer Cloud (puffercloud.dev, about Cloud), the optional hosted AI service run by the AI Puffer team. Connecting happens only after an administrator accepts the connection terms and clicks Connect. It sends the site’s Freemius install ID, the site address, a signed proof of the installation and the attempt’s support reference. Using its models sends the request content and model settings. Cloud passes each request to the AI service behind the chosen model (such as OpenAI, Anthropic or Google), through service providers such as OpenRouter where applicable, and returns the result and the credits used. Balance, model, checkout and disconnect requests send the site address and its Cloud credential; deleting the plugin also disconnects the site. While connected, the setup wizard sends your setup choices and completed steps. Connection diagnostics: each Connect attempt, including one that fails before account registration, sends its outcome, the step reached, error codes, safe network failure details, HTTP status, timing, software versions and a random support reference. Once an installation identity is available, the report also includes its site address and installation ID as unverified troubleshooting context. Your browser sends the report, so Cloud also sees the browser’s IP address and the site’s domain. Cloud keeps diagnostic reports for 30 days. The plugin keeps at most 10 local reports, removes those older than 30 days on the next attempt, and never reports on ordinary page loads. Terms · Privacy Freemius (freemius.com): account registration, email verification, licensing, and checkout for subscriptions and credits. Registration happens only when you choose it, for example by connecting Cloud or signing up for product updates in the setup wizard. It sends your name, email address and site address, plus basic plugin details such as the version. General usage tracking is turned off in the setup flow, and marketing emails are optional. The in-plugin pricing page, checkout and license activation also contact Freemius. An answer to the optional deactivation feedback form is sent to Freemius when the plugin is deleted and, unless you tick “Anonymous feedback”, registers the site. Terms · Privacy The AI providers and stores below are used only when you select one for a feature, or turn on a feature that uses it. OpenAI (api.openai.com): text, images, embeddings, vector stores, speech, transcription, web search, and moderation if you turn it on. A custom OpenAI-compatible base URL receives these requests instead. Terms · Privacy Google Gemini API (generativelanguage.googleapis.com): text, images, Veo video, speech, transcription, embeddings, File Search stores and Search grounding. Terms · Privacy Anthropic (api.anthropic.com): Claude text, image input and web search. Terms · Privacy xAI (api.x.ai): text, image input, web search and images. Terms · Privacy DeepSeek (api.deepseek.com): text. Terms · Privacy OpenRouter (openrouter.ai): text, images and web search, through the model vendor you pick. Terms · Privacy Microsoft Azure OpenAI (your Azure endpoint): text, images, embeddings and transcription. Terms · Privacy ElevenLabs (api.elevenlabs.io): reply text, when a bot reads replies aloud with ElevenLabs. Terms · Privacy Replicate (api.replicate.com): image prompts and settings. Terms · Privacy Pexels (api.pexels.com) and Pixabay (pixabay.com): your stock photo search terms. Pexels Terms · Privacy; Pixabay Terms · Privacy Openverse (api.openverse.org): when you ask a connected assistant to search stock images, your search terms and image filters go from your website to Openverse. No API key is needed. Results include creator, source and licence links; check the original source and retain required attribution. Searches are cached locally for one hour. Downloading a chosen image uses the existing media upload tool and contacts its image host. Terms · Privacy Pinecone (api.pinecone.io and your index host), Qdrant and Chroma (the address you enter), if chosen as your knowledge store: passages and embeddings of content you index, and search vectors for lookups. Pinecone Terms · Privacy; Qdrant Terms · Privacy; Chroma Terms · Privacy Hosted website management (mcp.aipuffer.org): only after you enable Manage my website and accept its disclosure, the bridge receives your site address, WordPress/plugin versions, Cloud account/site identifiers and public signing key to register the site. Assistant connections include the WordPress user identifier and permitted actions. Assistant requests and website results, including requested post text, pass through the bridge to your chosen assistant; draft content is sent to your WordPress site. Operational records contain identifiers, action names, status and usage, without post bodies or credentials. Recent operational history is searchable for 30 days; archives and completed admin command records are retained for 24 months, and unresolved cases are held until resolved. WordPress keeps its local operation records for recovery. Standard access requires WordPress approval for post publication; Trusted access can publish posts directly. Other plugins’ changes and deletions require approval at every level. Terms · Privacy. Your assistant provider’s own terms also apply. Webhooks: if you add webhook URLs in Settings → Developers, signed event data is sent to those addresses. Services used by optional upgraded features: Ollama sends requests only to the server you enter (default http://localhost:11434). RSS feeds, web pages and Google Sheets you choose as sources are read from their sites. Chatbot trigger webhooks and the Slack, HubSpot, Notion, Pipedrive, Zapier, Make and n8n integrations send the data you map to the app or address you connect. Source code The free edition’s original JavaScript and CSS, build tools and instructions are available in the public source repository. Version tags identify source snapshots matching the corresponding free release.
Newsroom OS – Google Discover, Google News, Schema & Editorial Workflow
Optimize articles for Google Discover, Google News, and web search. Add NewsArticle and LiveBlogPosting schema markup, Search Console traffic analytics, editorial assignments, and newsroom tools — directly inside WordPress. Works in Gutenberg and Classic Editor. Who is this plugin for? Perfect for: News websites and online newspapers Magazines and digital publishers Editorial teams and multi-author WordPress sites Sites targeting Google Discover and Google News traffic Publishers who outgrew generic SEO plugins but still use Yoast or Rank Math NewsArticle & LiveBlogPosting schema markup Structured data built for publishers — not generic blog schema: NewsArticle JSON-LD — @graph with Organization, WebPage, author Person, geo, about / mentions from entity tags LiveBlogPosting — breaking news mode with live H3 updates and auto-expire CollectionPage from story timelines; optional editor FAQ block (opt-in, off by default) Wikidata entity enrichment on tags (lazy, on assign) Richness presets (Minimal / Balanced / Full) — complements Yoast SEO, Rank Math, AIOSEO, SEOPress Google Discover & Google News optimization Discover Score — pre-publish checklist: hero 1200px+, 16:9, headline quality, topic/tags, OG image, timeliness Trend Radar — Google Trends + Google News RSS in the editor sidebar (localized by country) Breaking news mode — faster IndexNow/Google ping, LiveBlog schema, optional “developing story” badge Search Console clicks — real Discover and Google News traffic via GSC sync (separate from the editorial Discover Score) Newsroom OS vs Yoast SEO & Rank Math Yoast SEO / Rank Math excel at meta titles, descriptions, XML sitemaps, and general on-page SEO. Newsroom OS adds the newsroom layer they do not cover: Google Discover readiness scoring in the editor Google News + Discover Search Console analytics per writer NewsArticle / LiveBlogPosting publisher schema graph Editorial workflow — tasks, My Inbox, chief dashboard, Chief Insights Author performance coaching tied to real traffic Smart tags, duplicate-tag guard, archive internal links, GSC traffic-drop alerts Use both together — Newsroom OS reads your SEO plugin meta and adds publisher-specific signals. Update 2.9 — All Pro features included + install registry Newsroom OS 2.9 unlocks every Pro feature without a license key (GSC sync, Chief Insights, Author Performance, AI Assist BYOK, IndexNow, Auto-Timeline, task messaging, and more). A lightweight anonymous install/upgrade ping helps NewsOS count active sites. To restore paid license gating later, set NEWSAI_PRO_UNLOCKED to false in wp-config.php — all license code stays in place. Update 2.8 — Schema performance for megasites + nightly warm Newsroom OS 2.8 keeps JSON-LD on high-traffic sites without daytime CPU spikes: Lite Mode no longer runs the schema backfill cron, Background schema warm is optional, and Nightly schema warm runs migration only in an off-peak window (default 02:00–06:00, site timezone). New/edited posts still get full schema on save. On cache miss under Megasite mode, visitors get a slim NewsArticle graph while full caches backfill overnight. Also: Unicode-safe word counts (Greek), story-cluster accessibility contrast, and safer warm scheduling (hourly / 15-min only with backlog). Update 2.7 — Classic command center (stable) + parked React UI Newsroom OS 2.7 ships a premium classic admin redesign: vertical sidebar shell, bento Dashboard (Kanban-first), refreshed Chief Insights and Author Performance, and a cleaner Settings workspace. Classic is the default on all sites. An experimental React admin remains in the codebase for a future opt-in release — it is off by default; the Settings banner shows Coming Soon until the redesign is ready. Desk workflow: AI Content brief from Trend Radar now lives on the Dashboard (chief plans → assigns task → writer sees brief in the editor). The editor Radar tab stays lightweight (live trends/news + copy headline). Update 2.3 — Desk workflow + Chief command upgrades (Pro) Newsroom OS 2.3 adds brief → editorial task, AI desk overview (Author Performance), Kanban drag-and-drop, War room live KPIs (AJAX), CrUX / Core Web Vitals (PageSpeed Insights), GSC device channel toggle, and category-aware AI prompts. No database migration. Configure CrUX under Settings → Instant Indexing (PageSpeed API key). Update 2.2 — AI Assist SEO pipeline (Pro BYOK) Newsroom OS 2.2 upgrades AI Assist (Pro) for newsroom SEO: shared article analysis (keywords, entities, intent), stronger headline, optimize, and trend suggestions, meta description generation with one-click apply to Yoast SEO / Rank Math (or excerpt), smarter tag match, and Discover Score entity signals when analysis is saved on the post. Editor blocks (story timeline, key takeaways, FAQ) use semantic HTML, dedicated CSS (story-cluster.css, newsai-content-blocks.css), and no empty DOM nodes in published markup. Big update 2.1 — Chief Insights, AI Daily Brief, and indexing diagnostics Newsroom OS 2.1 adds Chief Insights (desk-wide KPIs, channel tables, top categories/articles), AI Daily Brief (Morning/Evening), and Print to PDF. Clear Last attempted vs Last successful diagnostics for Google Indexing API and IndexNow. Preview the product at newsos.io. Big update 2.0 — Google Search Console meets your newsroom Newsroom OS 2.0 connects your editorial desk to real Google traffic. Writers get Focus (article quality) and Discover readiness (distribution checklist) in the editor — separate from Search Console clicks (Discover, Google News, and web search). Chiefs see who publishes well and who earns traffic, in one dashboard. (From 2.9 onward, GSC sync and all former “Pro” modules ship unlocked in the same plugin — no license key.) Classic Editor & Gutenberg (Block Editor) Newsroom OS is built for real newsroom setups — including sites that still use Classic Editor, sites on full Gutenberg, and hybrid installs (block editor with the classic Tags metabox). Gutenberg / Block Editor Publication Readiness sidebar (Editorial + Discover twin gauges, Next Best Action) REST-aware duplicate tag guard (snackbar + “Use existing” / “Keep anyway”) Breaking news mode, live H3 update insertion via blocks Trend Radar, Smart Tags, AI Assist (BYOK) in the editor sidebar JS strings via wp_set_script_translations (el, en, de, es) Classic Editor Same Publication Readiness panel and scores (TinyMCE / #content hooks — no block editor APIs on classic screens) Classic Tags metabox duplicate guard (AJAX before the tag chip is added) Breaking mode, live updates, and smart tag insert via Classic UI Stability: Gutenberg-only APIs (wp.data.subscribe, getEditedPostContent) run only when block-editor-page is present — prevents Classic Editor blocks.min.js stack overflows on large sites Works with any theme (Astra, GeneratePress, Newspaper, NewsMag, Flavor, etc.) and alongside Yoast SEO, Rank Math, WooCommerce, and SureCart. WordPress 7.0: Tested through WordPress 7.0 (Classic Editor, Gutenberg, and wp-admin dashboard). Requires PHP 7.4+ (PHP 8.1+ recommended). Six pillars (2.0) 1. Discover optimization Publication Readiness — twin scores in the editor: Editorial (Focus) and Discover readiness. They are separate on purpose. Discover Score — Google-aligned checklist: hero 1200px+, 16:9, 300k pixels, alt; clear 55–70 headline (anti-clickbait); topic/tags; optional trend fit; saved OG image when available. Breaking news mode — one toggle: LiveBlogPosting schema, faster IndexNow/Google ping, optional “developing story” badge, auto-expire, live-update workflow (H3 headings). Trend Radar — Google Trends + Google News in the sidebar, localized per country. 2. Google Search Console traffic Service account sync — connect Search Console with webmasters.readonly; separate token from the Indexing API. Discover, Google News, and web search — per-post clicks, impressions, CTR, and average position stored in your database (wp_newsai_gsc_performance). Sync now in settings plus daily cron (04:00 site time) when GSC is enabled. Author Performance — real clicks per writer, top articles by traffic, coaching tab, 7d / 30d / 90d windows. Posts list — GSC (30d) column for chiefs (edit_others_posts). My Inbox — writers see their own GSC totals when sync is enabled. Clear labeling — Discover Score = editorial readiness; GSC numbers = measured Google clicks (not the same metric). 3. Semantic schema & entities NewsArticle / LiveBlogPosting JSON-LD — @graph with Organization, WebPage, author Person, geo, about / mentions. Entity tags — Wikidata Q-IDs, coordinates, entity types woven into schema; lazy enrichment when tags are used (1.9+). Richness presets — Minimal / Balanced / Full; optional articleBody modes (omit, full, first N words). Works with Yoast SEO & Rank Math — reads your meta; does not replace them. 4. Editorial workflow Tasks & assignments — deadlines, keywords, briefs, reference links. Dashboard — active / overdue / completed, KPIs, SEO health. Roles — chiefs see the full desk; writers see My Inbox only. GSC traffic-drop alerts — surfaces articles losing Search Console clicks vs the prior period. 5. AI newsroom (BYOK — your API key) OpenAI, Gemini, or Groq — your API key stays on your site; free tiers supported (Groq / Gemini). Discover-optimized titles (55–70 chars, five angles), targeted optimization hints, trend-aware keywords, AI Tag Match from your site taxonomy. Geo-aware language — answers in the language of your configured region (40+ locales). 6. Knowledge layer Smart Tag Suggester — weighted, confidence-scored tags from your taxonomy; one-click insert; duplicate-tag guard. Story timelines — archive clusters + internal-link blocks + CollectionPage schema. Archive link finder — semantic internal links from your own back catalog (accent-aware scoring). Author E-E-A-T — profiles, schema Person, performance analytics. UI in Greek, English, German, and Spanish (per-site language picker). See languages/README.txt for gettext catalogs. Everything included (2.9+) — 100% free, no license Newsroom OS 2.9 ships the complete newsroom stack in one plugin. No signup, no license key, no paywall inside WordPress. The only optional cost is AI Assist (BYOK) — you bring your own OpenAI, Gemini, or Groq API key if you want in-editor AI; Groq and Gemini offer generous free tiers. Editorial & Discover (free) Publication Readiness — Editorial score + Discover Score, Next Best Action, detailed checklists Breaking news mode — LiveBlog schema workflow with auto-expire Smart Tag Suggester — confidence levels, one-click insert, duplicate-tag guard Editorial dashboard — tasks, deadlines, keywords, briefings, Kanban, My Inbox Trend Radar — Google Trends + Google News in the editor Auto-story timeline + archive internal-link finder NewsArticle JSON-LD — entities, geo, E-E-A-T author signals, megasite-safe caching SEO auditor, 4-step wizard, 4-language UI (el, en, de, es) Analytics & desk command (free since 2.9) Google Search Console sync — Discover, Google News, and web search clicks per post and author Chief Insights — desk-wide KPIs, channel breakdown, top categories/articles, AI Daily Brief, PDF export Author Performance — team cards, GSC coaching tabs, 7d / 30d / 90d windows GSC traffic-drop alerts — in Chief Insights and weekly desk email Posts list GSC column — 30-day clicks for chiefs Seasonal radar — YoY keyword forecasting from GSC queries War room — live desk KPIs (AJAX refresh) AI & indexing (free since 2.9; AI is BYOK) AI Assist — headline suggestions, optimize hints, trend keywords, tag match, meta description, content brief, floating toolbar, GEO summary, auto alt text IndexNow + Google Indexing API — async pings on publish Task threads + email alerts — chief ↔ writer messaging per assignment Privacy: GSC data and AI cache stay in your database. Newsroom OS does not proxy article text or API keys through newsos.io. An anonymous install/upgrade ping (site URL, version, locale) helps NewsOS count active installs — no post content included. Feature reference (formerly “Pro”) 📊 Google Search Console integration Stop guessing which writers earn traffic. Newsroom OS pulls Search Analytics into WordPress: Three channels — Discover clicks, Google News clicks, and web search clicks (plus impressions, CTR, position where available). Per-post storage — map GSC page URLs to posts and authors; aggregate by 7 / 30 / 90 days. Settings → Indexing — paste your service account JSON, set the property (sc-domain:example.com or URL-prefix), test connection, Sync now. Automatic daily sync — when GSC is enabled with valid credentials. Privacy — data stays in your database; we do not proxy GSC through newsos.io. 📈 Author Performance Redesigned chief dashboard (Editorial Control → Author Performance): Your performance card for chiefs who also write — same expand/collapse UX as the team grid. Compact team cards — posts, Focus readiness, web clicks at a glance; expand for full detail. Tabs inside each author — Summary (editorial quality + categories + GSC totals), Articles (latest posts + top traffic), Coaching (Discover weak spots + action links). GSC coach tips on low click counts (hero image, headlines, timeliness). Search & sort — filter by name; sort by posts, web clicks, Focus, or A–Z. 7d / 30d / 90d — live AJAX refresh without reloading the page. Cached aggregates with smart invalidation on publish, task completion, score change, and GSC sync. 🧭 Chief Insights + AI Daily Brief (NEW in 2.1) Built for editor-in-chief oversight in one screen: Chief Insights tab — summary-first KPI strip + channel performance + top categories + top articles + top keyword/entity signals. Live desk data — real GSC-backed metrics when sync is configured. AI Daily Brief — one click executive summary with Morning mode (planning) and Evening mode (retrospective + next actions). Print to PDF — export the current chief view for standups and editorial meetings. Indexing diagnostics polish — Google + IndexNow now show Last attempted and Last successful URL/ping, plus health badge (Healthy / Degraded / Unknown). 🤖 AI Assist (BYOK – Bring Your Own Key) Bring your own API key for OpenAI, Google Gemini, or Groq. Newsroom OS never proxies AI traffic through our servers — your key stays on your site. 💸 Free AI tier supported: Groq offers fast hosted open-weight models with API keys from the same dashboard. Recommended default: openai/gpt-oss-120b; alternative: qwen/qwen3.6-27b. Deprecated Groq model IDs like llama-3.3-70b-versatile should be replaced. The settings page links directly to: https://platform.openai.com/api-keys (OpenAI) https://aistudio.google.com/apikey (Google Gemini — free tier available) https://console.groq.com/keys (Groq — free tier, no card required) You can write your first AI-assisted article with $0 of API spend. Direct in-editor actions: 🎯 Discover-optimized Title Suggestions — five alternative headlines per run (facts / impact / aftermath / authority / Discover hook), 55–70 characters, live validation. ✨ Targeted Content Optimization — non-rewrite suggestions per paragraph (readability / seo / eeat / structure / factcheck). 🔥 Trend-Aware Keywords — regional Google Trends fit with placement hints. 🏷️ AI Tag Match — semantic tag picks from your taxonomy pool. FAQ Block generation in valid FAQPage JSON-LD Key Takeaways block Per-user daily rate limits + transient caching 🌍 Geo-aware language output — 40+ locales; filter newsai_ai_geo_to_locale for custom multilingual newsrooms. 💬 Editorial Comms (Task Threads + Email) Two-way messaging thread per task — chief ↔ assignee, with unread badges Capability-aware …