Every recorded version, newest first, with what changed in each one.
Latest 3.9.09 versions
Undated · 9 versions
3.9.0
Latest
10 new3 improvements4 fixes
New
MCP Undo: every change an assistant makes can be undone in one click from the MCP tab, or just by asking, and the undo stops if someone edited the item since.
MCP approvals: approve deletions (or every site change) from the MCP tab, or keep assistants on drafts only.
MCP edits Gutenberg posts block by block, and turns Markdown into real blocks.
MCP works with your SEO plugin (Yoast SEO, Rank Math, All in One SEO, SEOPress, The SEO Framework, Slim SEO, SEO Engine) and with Advanced Custom Fields.
MCP edits pages built with Elementor and Beaver Builder (Pro), Divi, Avada, WPBakery and SiteOrigin.
MCP manages menus, CSS, site styles, block theme templates, classic widgets, WooCommerce coupons and variations, comments, draft previews and Redirection redirects, and reads form entries.
MCP checks pages as visitors see them, reads the error log and Site Health, and restores revisions.
Assistants get a short introduction to the site and starter prompts when they connect.
Auto Sync can target several embeddings environments, with the mwai_embeddings_sync_env_ids filter.
The chatbot launcher can reuse the AI avatar.
Improved
Redesigned MCP tab: the token setup lives in Connect, integrations show as chips, activity covers 30 days, and the Claude and ChatGPT directory links go through Meow Apps' connector.
MCP tools that can break things are flagged so assistants ask before running them, and tool descriptions are shorter.
Turning on Cross-Site for a chatbot also turns on Popup.
Fixed
Security: Image Studio now encodes the media ID in its requests, as extra hardening.
Trashing a post now deletes its vectors through each environment's own vector DB.
On mobile, the Glass and Timeless chatbots keep the input above the keyboard in long conversations.
A page that places the site-wide chatbot as a popup no longer shows it twice.
3.8.9
2 new3 improvements2 fixes
New
GPT-6.1 Terra model.
Restore link in Recent Changes, which opens the version from before an AI edit.
Improved
Connect card now offers a choice between a direct connection (recommended) and the Meow Apps relay, with labels under each choice explaining what it brings, and the address to copy now sits in the Copy step.
Connect card opens on the direct connection for Claude and ChatGPT, with the directory and the ChatGPT app one click away.
Image Studio's edit link now only accepts a media ID.
Fixed
Chatbot now explains why a file could not be sent to OpenAI, instead of showing OpenAI's "Expected an ID that begins with file" error.
OpenRouter model list now waits up to 30 seconds to load, for slow hosts.
3.8.8
6 new9 improvements11 fixes
New
MCP tab showing the week's activity, recent changes with links to each WordPress diff, connections, access and tools.
Settings backups with a restore in Maintenance, and settings now save only what changed.
Chatbots panel in Workspace to chat with the site's own chatbots, using their instructions, model and Knowledge.
Gemini 4 models such as gemini-4-argon now get vision, PDF, web search, Google Maps and thinking, as features follow the version number.
Keepalives for streams that stay silent from the start, so nginx no longer cuts them at 120 seconds.
The header shows the installed version and, when a newer one is out, how to update or renew.
Improved
The MCP connect card guides you through Claude's directory first, confirms when your assistant reaches the site, points ChatGPT users to the official AI Engine app with Developer mode as the direct option, says the connector stores no password and none of your content, and offers an optional, unticked Meow Apps news signup.
MCP stays off on updated sites that never stored the setting, only new installs get it on by default.
The MCP Connection Test warns when the host strips the Authorization header and gives the .htaccess line, and AI Engine reads that header from Apache when WordPress cannot see it.
The MCP post list accepts several post types or statuses at once, returns everything when asked for all posts, understands an author given by name, and the post count and post list tools return clearer data.
In the Knowledge tab, creating an environment moved into the environment dropdown and Create New became Add Manually.
New Meow Apps dashboard, from the latest shared common code.
The admin now works on tablets, with the header, columns and Chatbots toolbar no longer breaking.
The header lights the button of the screen you are on.
Various labels and links, including the Dashboard footer now pointing to the Getting started guide and the Discussions setting pointing to the Discussions tab.
Fixed
Security: Shop Managers can no longer edit administrator accounts through MCP, and the plugin and theme file tools now follow DISALLOW_FILE_EDIT.
Security: Guests can no longer reset their usage limits with a fake IP header, while proxies and Cloudflare are still recognized.
Security: Chatbot replies now drop embedded frames and forms, and guest uploads are capped at 20 per hour.
Chatbots keep web search, MCP servers and their reasoning settings after calling a function, which also stops Claude's "Invalid signature in thinking block" error.
Chatbots keep their Knowledge context after calling a function on every provider, including Google, OpenRouter and other compatible providers.
Syncing a large OpenAI Vector Store no longer times out, each step stops after about 25 seconds and the next one continues from there.
The chatbot preview in the Chatbots tab shows the conversations held before 3.8.7 again instead of starting empty.
Content Studio no longer shows each section heading twice when the model repeats it.
MCP tools now accept text request ids as the JSON-RPC spec allows.
Re-setting a featured image through MCP no longer fails, terms can be removed, and featured image errors say why.
Admin scripts refresh when only the UI library changes instead of staying cached.
3.8.7
6 new9 improvements21 fixes4 changes
New
Explore AI Engine, a ten-goal tour from the dashboard that follows you as a slim bar onto each screen, with Knowledge guidance in the chatbot settings.
Workspace mobile app on the dashboard, with QR code, store links and a status line once a phone is paired.
Sakana AI (Fugu and Namazu) as a native provider, hidden until switched on with MWAI_SAKANA.
Read-only access for Claude and ChatGPT MCP connections, with a switch per connection in Connected Apps.
Clean Up button in Knowledge for rows of deleted environments and orphans.
The mwai_openai_body filter, to set any OpenAI parameter AI Engine leaves to the defaults, like high quality for the chatbot's image tool.
Improved
Redesigned the dashboard with a Your AI card for models and providers, a This week chart compared with last week, shuffleable idea tiles and per-company AI visibility bars.
Glass now takes its colors from the lights you pick for the visitor's bubbles, the launcher and the accent, and the theme editor preview opens with a sample conversation that is never saved or sent to the AI.
New popup chatbots now open in the macOS frame and header, while existing chatbots keep the frame they already have.
The MCP screen now shows how many tools each set brings and how many AI Engine has in all.
MCP deletes without force now always go to the trash (products, events, media and comments too) or refuse clearly, and deletes that would be permanent ask the AI to check with the user first.
The Settings screen, dashboard cards and feature showcase are now translatable, and Knowledge, Privacy First and Content Studio wording is clearer.
Every Pro link in the admin now points to the AI Engine page and is tagged, Pro-locked buttons open it instead of doing nothing, and Pro chatbot modes show their Pro pill for free users.
The Pro update check now reports whether MCP was used in the last 30 days and how many phones are paired.
The chatbot preview in the admin now keeps its own history, and the dashboard uses the same spacing as the other tabs.
Fixed
Security: Voice chatbots now only run functions attached to their own chatbot.
Security: Editors on MCP can no longer publish other content types or file references through a decoy fields object, and Public API tools and MCP servers stay restricted to administrators.
Security: Cross-Site requests are now matched on their real REST route, so other REST calls can no longer skip the nonce check by mimicking its URL.
Security: The Gemini file upload now verifies TLS.
Security: The MCP sign-in page can no longer be framed by another site, shows where the app sends you, and warns when an app calls itself Claude or ChatGPT without being it.
Insights logs work again on sites whose logs table predates 2.5.6, since tables now get their missing columns on update.
Chatbots and forms no longer stay blank when one of their settings contains entity text like ".
Posts containing an escaped "
AI Forms in chatbot mode now use the chatbot's own Max Tokens, model and temperature instead of hidden form values that cut replies at 4096 tokens.
Messages, uploads, stopping a reply and loading a conversation now follow the right chatbot's discussion when a chat id belongs to another chatbot.
Claude no longer rejects the answer after a function call when Knowledge is on (Invalid signature in thinking block).
When a thinking model spends its whole Max Tokens reasoning, the reply now says so instead of showing the raw reasoning as the answer.
Knowledge no longer retries rows of deleted environments, and large OpenAI Vector Stores sync page by page instead of timing out.
MCP connections no longer drop when a client revokes the token it just replaced after a refresh, and sign-in responses say which site answered.
Links in the GDPR consent text are now clickable instead of shown as raw text.
The site-wide popup chatbot no longer leaves a screen of blank space at the bottom of the page on themes that give the body a height.
The user's IP is shown again under their name in Discussions and Insights for logged-in users, and in the discussion details.
Chatbot names with & no longer show &.
Admin screens no longer load blank, and the header now shows the Free and Pro tiles.
Links on other plugins' screens are no longer tagged as AI Engine traffic.
OVHcloud and Sakana AI now follow one rule: hidden until switched on with MWAI_OVH or MWAI_SAKANA, and always shown where an environment of that type exists.
Changes
🎵 Discuss with others about Ai Engine on [the Discord]
🌴 Keep us motivated with [a little review here]
🥰 If you want to help us, check our [Patreon]
🚀 [Click here]
3.8.6
1 new7 fixes
New
The MCP screen's Claude step now opens Claude with this site's connector already filled in, so there is nothing to copy and paste.
Fixed
Security: Editors can no longer change the title or excerpt of posts they are not allowed to edit (coupons, menus, templates) through the AI helpers, and the other helpers taking a post ID now check permissions too (reported by Anatoly Gorbounov, x1will, via WPScan).
Hotfix: Chatbot replies no longer crash on servers with an older cURL, such as Xserver, because of a constant added in 3.8.5.
Messages like "0" or an image alone no longer fail the reply when Knowledge is on.
Chatbot saves silently refused by the database now show an error.
A Max Height set without a unit is now treated as pixels.
AI Forms set to a specific environment with the Default model now work instead of failing with "Model not found".
Settings changed while a chatbot reply is running are no longer undone when that reply saves its usage stats, which could revert a refreshed model list.
3.8.5
5 new6 improvements7 fixes
New
GPT Realtime 2.1 and GPT Realtime 2.1 Mini for voice chatbots.
Claude Haiku 5.5.
Notice on new sites to turn on the MCP module in one click.
Warning in Settings and the chatbot editor when Knowledge has no embeddings model to search with.
Optional comment_parent on the MCP wp_create_comment tool, to reply inside a comment thread.
Improved
Playground presets are now a system prompt with parameters: picking one no longer fills the message box or resets the models being compared, the message box sits on top while the conversation is empty, and reply stats read plainly.
The Knowledge Export button is now always visible under the table, instead of only in Expert Mode.
AI Forms in chatbot mode now use the chatbot's functions and MCP servers.
Gemini 3.6 and newer no longer receive the temperature setting, and Gemini 3 and newer no longer receive a thinking budget.
The Perplexity connection check now uses the models endpoint, and shows Perplexity's own error when the key is wrong.
Streamed replies stay connected through long silent phases such as Code Interpreter or slow tool calls, and stop calling the AI provider when the visitor leaves.
Fixed
Security: MCP no longer returns API keys or tokens when reading WordPress options.
Security: Uploaded files can no longer carry metadata AI Engine uses internally, and the expired files cleanup only deletes the expired file itself.
ChatGPT connections no longer fail with a missing resultType error when MCP requests use the 2026-07-28 format.
ChatGPT and Claude can now sign in to MCP on servers that block paths containing /.well-known/ under /wp-json/, like many nginx setups.
Connecting ChatGPT or Claude no longer fails with an expired code right after approval on sites with a persistent object cache.
Claude Sonnet 5.5 cache reads now cost $0.10, as Anthropic charges.
AI Engine no longer loads the block editor scripts on Classic Editor screens, which broke plugins like HTML Editor Syntax Highlighter.
3.8.4
8 new8 improvements4 fixes
New
Claude Opus 5.5 and Claude Sonnet 5.5.
GPT-6 Sol, GPT-6 Luna and GPT-6.1 Sol.
MCP tools to create, update, resolve and delete editor Notes, attached to the right block.
Gemini image sizes 512px, 1K, 2K and 4K, from the mwai_image MCP tool or the imageSize query param; models limited to 1K keep their default.
The mwai_image MCP tool now accepts an environment, a model, a quality and a file name per call, names files after the title, and records the model used.
An mwai_generated meta on every image AI Engine saves to the Media Library, set before wp_generate_attachment_metadata runs.
REST routes to enable MCP and check its status.
The mwai_anthropic_body filter, to tune the Anthropic request such as the web search limits before it is sent.
Improved
Knowledge now keeps at least 16,384 characters of each post instead of the old 4,096 default kept on older sites, and long posts are not re-synced after the update.
Site search with Knowledge or AI keywords skips bot and spam searches, and the same search is only paid for once a day.
The MCP Plugins and Themes options now say they can read and edit files, and the tool sets remind you to refresh the tools in ChatGPT or Claude after a change.
Workspace conversations continue on the model of their last reply, a new chat starts on your default model, and a default the site cannot run falls back to the first model.
Pinecone and Qdrant now keep each chunk's text and post ID, so Pull All rebuilds a working Knowledge base after a site move.
Current Claude models can now be picked as the default model for JSON queries.
Marked Claude Sonnet 4.5, GPT-5.4 Nano and GPT-5.1 as deprecated.
The review notice now shows two days after the first successful MCP tool call.
Fixed
Security: The search endpoint used by the Search settings screen is now limited to administrators.
Qdrant and Chroma server URLs pasted from the browser, like :6333/dashboard#, now work instead of sending every request to the dashboard page.
Streamed replies are no longer cut after 5 minutes while still arriving, only when the connection goes silent.
The JSON helpers used by Meow Apps Health and add-ons now work on Claude and Gemini sites with default settings.
3.8.3
7 new5 improvements16 fixes
New
Abilities to MCP, so tools registered by plugins through the WordPress Abilities API work over MCP and in Workspace, appear in MCP Logs, and are labelled Ability or Native in the renamed MCP Tools & Abilities list.
Cached tokens are now counted and priced at the cached rate for OpenAI, Anthropic, DeepSeek, OpenRouter and Claude Fable 5, and Insights shows cached tokens per query.
OpenAI requests carry a prompt cache key per chatbot, so repeated prompts and tools are billed at the cached rate.
A reply cut by Max Tokens now ends with a short notice instead of stopping mid-sentence, in chatbots and forms.
A chatbot limited to one file shows the attached file name and a remove button before sending, like the multi-file mode.
Prices for Gemini text models, so their queries are costed and dollar limits apply to them.
The mwai_openrouter_body filter, to add any OpenRouter parameter to a request, such as provider routing.
Improved
The MCP settings were reorganized, with the Connect card filling its block, Connections and MCP Settings under it, tool sets and WordPress abilities under the tools list, and the bearer token masked in the Claude Code commands.
The Files Manager lists files uploaded by every user and guest instead of only your own, and expired files are cleaned up every hour instead of once a day.
New chatbots start with their texts in the site language instead of always in English, existing chatbots are untouched.
OpenRouter requests are attributed to AI Engine instead of each site, and the existing filters still let a site credit itself.
Workspace chats are grouped under Today and Yesterday in the user timezone instead of UTC.
Fixed
The Editor Assistant no longer fails with Feedback session expired after applying edits, saved queries no longer store API keys, and it now sees paragraphs edited after the page loaded.
Claude, Gemini and other non-OpenAI chatbots on an OpenAI Vector Store now receive the matching knowledge instead of an empty context.
Non-streamed errors from OpenRouter, Mistral, OVH and custom providers are shown as a message instead of raw JSON.
Gemini replies cut by Max Tokens end with the notice instead of an empty-reply error, and Gemini errors wrapped in a list are readable.
Picking an environment without a model now asks for one instead of sending another provider's default model and failing with a 404.
An AI Form repeated on a page answers in its own copy, and a streamed form error no longer floods the PHP log.
GPT-6 Astra can generate images, its model entry was missing the image generation tool.
Uploads that fail now say why, with the size limit when the file is too large, instead of an unknown error or a wrong invalid file type message.
An empty completion request is refused before reaching the provider, and Suggest Synonyms without a selection returns a clear error instead of a fatal.
MCP tool failures, like a missing post, reach the AI as a readable tool error instead of a protocol error.
ChatGPT MCP connections no longer drop when several of its sessions share one connection, and revoking an app disconnects all of its sessions at once.
When AI Engine manages the WordPress Connectors, an empty key no longer wipes a saved one, and keys already saved in WordPress are reused.
Syncing a single post that gets skipped (ignored, language filter, or filtered out) now says why instead of doing nothing.
Copying a masked MCP command by hand copies the real token, and Knowledge asks to pick an environment instead of showing the first-run screen.
On touch screens the Stop button works with a single tap, while a quick double tap on Send still cannot cancel the message.
Two PHP warnings on every logs request when Insights is off.
3.8.2
1 new1 improvement3 fixes1 change
New
Connect an AI assistant card in the MCP settings, which checks the host, guides you through Claude, ChatGPT or Claude Code step by step, and confirms the connection works.
Improved
The MCP settings and the readme now state that connections are direct, with no relay, no fee or cap, and credentials staying on the site.
Fixed
MCP connections no longer break when a client, such as ChatGPT, reuses a refresh token that was already replaced.
Chatbots with streaming disabled no longer slide the window up when a reply starts, which hid the Glass header and could make the page jump.
Add-ons now keep their own visitor params as sent; only denied keys are matched by their canonical name and only messages is folded into one key.
Changes
For older releases, see [changelog.txt]
Changelog published by the developer, in their original language.
