What's New

Add any remote MCP server as an org-public resource, then attach it to scenarios by id — not just the curated catalog. Paste a Streamable HTTP URL, set Header: {{ KEY }} rows, and pick which resource servers a scenario uses.

How it works:

• Scenario Studio → Form → MCP Integrations → Add your own MCP server
• The URL lives on the org resource; scenarios store only custom:<oid>
• Secret values stay in Settings → API Keys — each runner injects their own key
• Catalog servers (GitHub, DeepWiki, Context7, Linear, Tavily, Exa) sit next to your custom ones

Works on Ocean and Landmass (including phone). HTTPS only. OAuth for custom servers comes later.

Every scenario now keeps a full version history. You can see past versions, compare them, and keep exactly one active. This was a long ask from teams who test variations off customer feedback and keep evolving their scenarios.

What you get:

• Full history — every saved change is kept as its own version
• Compare — see exactly what changed between two versions before you commit to one
• One active version — the version learners run is always unambiguous, no matter how much you experiment

Connecting the ToughTongue MCP server no longer requires creating an API key. Add the server URL in Cursor, Claude Code, Codex, Copilot, Windsurf, Gemini CLI, claude.ai, or ChatGPT, and sign in with your ToughTongue account — a browser window opens, you approve access, and you're connected.

What changed:

• OAuth 2.1 sign-in with dynamic client registration on the hosted MCP server — add the URL with no credentials and the client walks you through the rest
• No key management — the sign-in issues a token behind the scenes; view or revoke it anytime at Developer settings
• PAT bearer auth still works for servers, CI, and headless agents

Endpoint: https://api.toughtongueai.com/api/public/mcp

Practice sales calls, interviews, and coaching — now straight from Claude, Codex, or Cursor. One plugin install gives your coding agent two things: Skills it loads automatically, and an MCP server it can call directly.

Skills shipped today:

• scenario-creator — builds a practice scenario from a brief, URL, or call transcript
• scenario-refiner — diagnoses and fixes a live scenario from real session evidence
• session-analyst — aggregates scores and weaknesses into coaching reports
• getting-started — post-install onboarding

MCP server: 27 tools over the ToughTongue API — scenarios, sessions, analytics, SIP calls, meeting bots, and collections. The agent can act, not just advise.

Works with Claude Code, Codex, Cursor, Copilot, Windsurf, Gemini CLI, or any agent that supports Agent Skills + MCP — and composes with whatever else is connected, e.g. "pull the last 3 lost deals from Gong and build a pricing-objection scenario."

Agent Desktop now ships full purpose-built apps, not just markdown pages. Assign your agent a real application, and it drives that app live to make its point visually instead of just in words.

Apps available today:

• Brainstorm Board — the agent captures and clusters ideas as you talk through a problem
• Prioritization Framework — stack-rank options against your criteria with a rich set of visualizations

How to use it:

1. In the Advanced Scenario Editor, enable Agent Desktop and pick the apps to expose
2. Tell the agent when to reach for each app in your scenario instructions
3. During the session, the agent launches, focuses, and edits the app live — building on it turn after turn instead of starting over

More apps are on the way — every registered app gets the same launch/read/write toolset for free.

Your scenario can now learn from its own sessions. Set a learning config once, and after each session the agent conservatively checks the transcript for scenario-design issues — confusing setup, unclear instructions, missing context — and flags only what's worth fixing.

What you configure:

• A learning prompt describing what to watch for
• A session analysis limit — how many future sessions get analyzed

What comes back: a summary, the specific issue, a transcript quote as evidence, and a suggested fix — stored per session, never invented.

How to use it:

1. In the Advanced Scenario Editor's Analysis tab, turn on Scenario Learning and set your prompt + limit
2. Sessions run as normal — learning notes accumulate quietly in the background
3. Open Edit with AI and ask it to fix the flagged issue — e.g. "the onboarding agent ends calls too early, diagnose and fix it" — it pulls the notes and applies a surgical update

Conservative by design: normal learner weakness is ignored, only true scenario-design issues get surfaced.

Truely, our invisible-to-screen-share desktop assistant, is now available on Windows. Run scenario-configured AI copilots live during your calls — sales assistance, meeting assistance, interview coaching, and more.

What you get:

• Sales assistance — live objection handling, talk tracks, and CRM-aware prompts mid-call
• Meeting assistance — real-time notes, follow-ups, and action items as the conversation happens
• Invisible to screen share — the overlay stays hidden from screen capture and recordings
• System audio capture — Truely hears both you and the other side via Windows loopback

How to get it:

1. Open the Truely page and download the Windows installer
2. Run the setup and sign in
3. Pick an assistant and launch — the overlay floats above any meeting app

Available now alongside the macOS build (Apple Silicon and Intel).

Your Agent can now build and update a shared visual workspace during a live session. Agent Desktop opens inside the artifact panel and gives the Agent two apps: Canvas for markdown pages and Custom App for richer React/TypeScript interfaces.

What you can create:

• Canvas pages — briefs, agendas, rubrics, debriefs, tables, and action plans under /canvas/pages/
• Custom Apps — dashboards, product mockups, visual reports, comparison boards, and data views under /custom-app/
• Editable files — inspect and update app files live through FileExplorer while the session continues

How to use it:

1. In the Advanced Scenario Editor, enable Agent Desktop in Tools
2. Tell the Agent when to use Canvas versus Custom App in your scenario instructions
3. Start a session — the Agent can open Agent Desktop, create files, and revise the visual artifact as the conversation evolves

Agent Desktop is best when learners need a persistent board, report, checklist, or interactive visual that improves during the session instead of disappearing after one response.

Your AI agent can now use external services — GitHub, documentation portals, project trackers — as live tools during a conversation. Mid-call, the agent pulls code, looks up docs, or creates issues without you lifting a finger.

Available today:

• GitHub — browse repos, read files, search code
• DeepWiki — query open-source documentation live
• Context7 — look up library APIs and code examples

How to set up:

1. In the Advanced Scenario Editor, expand Integrations and pick the services you need
2. Add your API key once in Settings → API Keys
3. Start a session — the agent calls those services as needed

Works across browser, phone calls, and meeting bots. Keys stay server-side — never exposed to the client.

Connect your own model provider keys to Tough Tongue AI and run practice on your own quota. Bring a free Gemini key from Google AI Studio, or use your existing OpenAI and other provider keys.

What you get:

• A flat 1 cent per minute platform fee, which is negligible enough to feel like unlimited practice
• Support for Gemini, OpenAI, and other model providers
• Avatar orchestration with Anam and other providers, all run through our platform

How to set up:

1. Open Settings and go to the API Keys section
2. Add a key from Google AI Studio (Gemini), or paste an existing OpenAI key
3. Select your provider when creating or editing a scenario
4. Start practicing on your own quota

Availability: BYOK is a Premium feature, included at $20/month.

Your inbound SIP calls can now fetch caller context before the AI agent picks up. The caller hears normal ringing while your endpoint is called — the agent joins with full context, no dead air.

Without hook: "Hello, how can I help you?" With hook: "Hi Ravi, welcome back. I see you're on the Premium plan."

How to set up:

1. Create a custom function with type Pre-Connect, pointing at your context endpoint
2. Select it in the scenario's Pre-Connect Hook dropdown
3. On inbound call, your endpoint receives SIP metadata (caller number, trunk ID, headers)
4. Return user_name, user_email, and/or dynamic_vars — injected into the agent prompt as {{variables}} from the first utterance

Fail-open: If your endpoint fails or times out, the call still connects without enriched context. The caller is not dropped.

Use cases: CRM personalization, resolving real caller numbers in forwarded call chains, and multi-tenant routing where one number is shared across businesses.

Get notified the moment a session moves through its lifecycle. Tough Tongue AI now POSTs signed HTTP events to any endpoint you register — no polling, no delays.

Six granular events:

• session.started — session created via browser, phone call, upload, or meeting bot
• session.completed — session finished and recording is available
• session.analyzed — evaluation results and next steps are persisted
• session.extracted — extraction results are persisted
• session.terminated — session force-terminated by the system
• post-session.done — configured post-session work is complete and persisted

How to set up:

1. Go to Developer → Webhooks → Add webhook
2. Paste your endpoint URL and select which events to subscribe to
3. Copy the signing secret — verify every delivery using Standard Webhooks

Delivery guarantees: Each event is delivered once with a 10-second timeout. Use the id field in the payload for idempotency. Delivery logs (status code, payload) are stored for 7 days per webhook.

Perfect for: pushing session results to your CRM, triggering Slack alerts when scores are ready, piping extracted variables into your database, or monitoring call quality in real time.

Your scenarios can now answer the phone. Connect a SIP trunk, link a phone number to a scenario, and any caller dialing that number will be routed straight into a live conversation with your agent — with full transcript, recording, and analytics, just like outbound calls.

How it works:

• Open Settings → Phone Calls and add a SIP trunk with type Incoming
• Pick the scenario that should answer the call and (optionally) restrict to specific DIDs or carrier IPs
• Point your carrier (Twilio, Telnyx, Plivo, Exotel, IVR Solutions, Wavix) at the LiveKit SIP URI we generate
• Done — every inbound call now lands in that scenario

What you get on every inbound call:

• Full transcript and audio recording stored on the session
• Caller's phone number captured automatically (searchable in analytics)
• Same evaluation, variable extraction, and post-call workflows as web sessions
• Per-trunk DID allowlist + IP allowlist for security; auth username/password supported when your carrier requires digest auth

Where to see it: Open any scenario's Phone Call Integration page — when an incoming trunk is linked, you'll see the dialable numbers right alongside the outgoing call controls.

Introducing Truely — a real-time desktop meeting assistant that runs as an invisible overlay during your live calls. Download the Mac app, pick an assistant (or build your own), and get silent, text-only coaching while you stay in control of the conversation.

What makes it different:

• Invisible: Undetectable during screen share. A low-latency floating overlay delivers insights as fast as you need them
• Yours to configure: Upload your knowledge base, write the system prompt, attach product docs, sales playbooks, or competitive intel. The assistant uses your data in real time — not a generic model
• Action-capable: Create Linear tickets, read from Jira, pull CRM data, and execute workflows — during the meeting, not after

6 featured assistants ready to go: Sales calls, cold calls, PM syncs, hiring interviews, customer success QBRs, and investor pitches. Use them as-is or remix them into your own custom assistant with your own knowledge base and tools.

Your scenarios can now automatically extract structured variables from the conversation transcript after each session ends. Instead of a written analysis report, get clean, queryable fields your CRM, dashboard, or pipeline can consume directly.

How it works:

• In the Advanced Scenario Editor, open Session Analysis and enable Variable Extraction
• Define each variable: give it a name, a description that tells the AI what to look for, and a type (text, number, boolean, list, or date)
• Run a session — extraction runs automatically when the call ends
• View structured results in the session detail page

Example — Sales call:

Add three variables and every completed call returns: lead_name (text), city (text), conversion_likelihood (number, 1–10), and conversion_sentiment (text — positive / neutral / negative). That's enough to prioritize your follow-up queue without reading a single transcript.

When to use extraction vs. rubric analysis:

• Extraction → when you need structured fields: CRM sync, lead routing, dashboards, batch reporting
• Rubric analysis → when you need prose feedback (coaching, performance review, training assessments) or multimodal / audio analysis where the full recording matters
• Both can run on the same scenario simultaneously.

Your AI agent can now take real-world actions during live conversations using pre-built integration templates. Pick a template, paste your API key, and your agent is ready to go.

What your agent can do:

• Send emails via SendGrid
• Send Slack messages or SMS/WhatsApp via Twilio
• Create issues in Jira or Linear
• Book meetings on Cal.com
• Append rows to Google Sheets
• Trigger any webhook

Each template comes pre-configured with the correct HTTP method, URL, headers, and parameter schema. Just replace the placeholder API key with your own and attach the function to your scenario.

Our AI phone calling experience has gotten a serious upgrade, especially for Indian geographies.

What we shipped:

• Colocated infrastructure: LLM, STT, and TTS colocated for key geographies to cut latency
• Improved multilingual handling: Hinglish, regional languages, and code-switching flow naturally
• Human-like speech: Filler words and speech patterns like "haan ji," "umm," and natural pauses make calls sound real
• Interrupt-aware pipeline: Handles constant acknowledgements ("haan ji," "achha") and background noise without freezing or restarting
• Audio-to-audio pipeline: For use cases like data collection where STT-LLM-TTS loses information

If you've tried building voice AI for Indian phone calls, you know the problem — noisy networks, chatty conversations, and every other sentence has a "haan ji" thrown in. Most voice AI treats that as an interruption. We taught our agent the difference between someone acknowledging and someone actually trying to speak.

This comes at no additional cost. Bring your existing SIP config and plug it in.

Your AI agent can now embed a form (Google Forms, Typeform, JotForm) and display it to the user during a live call. The user fills it out inline without leaving the conversation. Great for feedback, reflection exercises, and post-session surveys.

Two ways to trigger it:

• AI instructions: Tell the agent to open the form after covering specific content (e.g. "Once you finish the debrief, show the feedback form")
• Conductor message: Schedule a timed message that instructs the agent to show the form at a specific point in the session

Setup: Enable the Form tool in the Advanced Scenario Editor and paste your form's embed URL.

Your AI agents can now join private and protected Google Meet rooms. The bot joins as a signed-in Google account, so it works with any access setting.

How it works:

• Open meetings: Bot joins automatically, no extra steps
• Private meetings: Two options:
  • A participant clicks "Admit" when the bot knocks
  • Add ttai@toughtalkai.com as a calendar attendee and the bot auto-joins without knocking

The auto-join approach is recommended for agent-led scenarios (screening interviews, coaching, assessments) where the bot must operate independently.

You can now bring any voice — including your own cloned voice — directly into your scenarios using Cartesia TTS.

How to use:

1. In the Advanced Scenario Editor, select Cascade as your AI model
2. In the TTS Provider section, switch to Cartesia TTS
3. Paste your Voice ID from play.cartesia.ai/voices

Cloned voices are fully supported — just paste the voice ID of your cloned voice the same way. Leave the field empty to use Cartesia's default voice.

You can now switch between dark and light themes! Toggle your preferred theme from the settings to match your environment and comfort.

Your AI agents can now call external APIs in real-time during live conversations. Define custom HTTP functions — and the agent intelligently decides when to invoke them based on conversation context.

What this unlocks:

• Book meetings — Agent checks availability and books a Calendly slot mid-conversation
• Update your CRM — Log call outcomes, update deal stages, or create contacts in HubSpot/Salesforce while still on the call
• Send emails & SMS — Trigger a follow-up email or confirmation text without leaving the conversation
• Fetch live data — Pull order status, account details, or inventory in real-time to answer customer questions instantly
• Trigger workflows — Kick off n8n, Make, or Zapier automations from a voice conversation

Example — Calendly Booking Agent: Create a function named book_meeting with method POST pointing to your Calendly API endpoint. Define parameters like { "invitee_email": "string", "date": "string", "time_slot": "string" }. The agent will collect the prospect's preferred time during the conversation, call your endpoint to book the slot, and confirm the booking — all without the user lifting a finger.

AI agents can now open URLs and navigate web applications in real-time while talking to the user. The agent controls a cloud browser — clicking, scrolling, filling forms — and the user watches everything happen live.

Use cases:

• Product onboarding — Walk new users through your app step-by-step while explaining each feature
• Training & enablement — Demonstrate internal tools, dashboards, or workflows hands-free
• Customer support — Show users exactly how to complete a task instead of describing it
• Sales demos — Give interactive product walkthroughs during live conversations

Supports authenticated sessions — pre-configure login credentials so the agent opens your app already signed in. No manual setup needed for each demo.

Analyze conversations from any source with your scenario's rubric! Upload transcripts or audio recordings to get AI-powered feedback using your custom evaluation criteria.

Two ways to use:

• UI: Scenario page → Sessions tab → Upload button (up to 10 sessions at once)
• API: POST /api/public/v2/session with transcript or recording_url

How it works:

1. Submit transcript text directly OR provide an audio URL
2. Analysis runs against your scenario's rubric
3. View results on the session analysis page

Perfect for analyzing sales calls recorded elsewhere, evaluating interview recordings from external tools, or batch-processing historical conversation data.

Use private Google Slides presentations without publishing them to the web! Simply share your presentation with our service account email, and the agent can display slides during conversations.

Setup: Paste your regular Google Slides URL → Share with service account → Verify access → Done.

Perfect for confidential training materials, proprietary content, or presentations you don't want publicly indexed.

Sign in with Outlook, Hotmail, and other professional email domains. This makes it easier to collaborate with teammates and bring your organization onto Tough Tongue AI.

Our agents can now capture snapshots from your webcam and screen to provide real-time visual feedback. Enable these tools in scenario studio to create immersive coaching experiences.

Webcam Support - Observe body language, posture, eye contact, and presentation style. Great for public speaking and interview presence coaching.

Screen Support - See what users are working on to guide them through problems. Perfect for tutoring, design reviews, and product demo practice.

AI agents can now generate presentation-style slides with image on one side and bullet points on the other.

This image + bullets layout has been used in business and education for decades — borrowing from this familiar pattern makes the output feel polished rather than experimental. Great for coaching and training where visuals reinforce learning.

AI agents can now trigger floating emoji reactions (⭐ 👍 🎉 ❤️ 🔥) to celebrate correct answers, milestones, or achievements during sessions.

Enable the Emoji Reactions tool in your scenario settings to let your AI coach provide instant positive reinforcement.

Resume interrupted sessions directly within embedded scenarios! Add ?restoreSession=SESSION_ID to your iframe embed URL to restore a previous session's conversation context. The AI picks up exactly where the session left off, maintaining full transcript history.

Perfect for handling connection drops, long-running training sessions, or allowing users to continue practice across multiple sittings.

Distribute your scenarios with powerful new share link generation! Create either general reusable links or single-use links with automatic expiration. Configure pre-check screens, user info prompts, and access controls - all from a simple dialog.

Key Features:

• Single-Use Links - Perfect for interviews and assessments with automatic expiration after one session
• General Links - Reusable links for ongoing practice and training platforms
• Token-Based Security - Private scenarios automatically include 24-hour access tokens
• Customization Options - Toggle pre-check screens and user info collection

Perfect for conducting interviews at scale, distributing practice sessions, embedding scenarios on websites, and deploying agents to end-users with controlled access.

Create stunning custom avatars with full creative control. Unlike preset Live Avatars, Hybrid Avatars let you upload your own thinking and speaking video states - perfect for branded characters, 3D isometric designs, or any visual style you can imagine.

Key Features:

• Full Creative Control - Upload custom thinking and speaking videos; no preset limitations
• Lifelike Animations - Realistic lip-sync and smooth state transitions
• Any Design Style - Support for 3D renders, illustrated characters, or realistic footage

Connect your phone numbers to Tough Tongue AI so your voice agents can make real phone calls. Perfect for sales teams, customer support, appointment scheduling, and training scenarios.

Setup takes 10-15 minutes via SIP trunk configuration in your scenario's Phone Call settings.

Add realistic, lifelike avatars to your scenarios with synchronized lip movements that match the AI's speech in real-time. This feature creates a more immersive and natural conversation experience.

Key details:

• Only available with Landmass models (select from the AI model dropdown)
• Additional cost: +0.5x minutes usage

Transform how your team uses Tough ongue AI with our comprehensive organization management features. Perfect for training teams, educational institutions, and enterprises.

Key Features:

1. Organization Setup - Create your organization with custom branding and subdomain URLs.

2. Team Member Management - Add members individually or in bulk with role-based permissions (Admin vs Member). Admins can manage minutes and members.

3. Granular Access Control - Google Drive-style permissions for scenarios with three access levels: View only their sessions (least access), View all sessions (audit access), or Edit scenarios and view all sessions (full access).

4. Course Organization - Group multiple scenarios into courses for better organization.

5. Centralized Billing - Manage organization minutes, track consumption, and access billing history.

6. Creator Analytics - Track personal and organizational usage with creator analytics.

7. Developer Integration - Create API keys and integrate scenarios into your applications.

Customize the AI avatar in iframe embeds using simple query parameters! Override the default scenario avatar by adding ?avatarUrl=YOUR_IMAGE_URL or ?avatarId=NUMBER to your embed URL.

Available options:

• avatarUrl - Use a custom avatar image from any URL
• avatarId - Select from numbered default avatars

Perfect for matching the avatar to your brand identity or context when embedding scenarios across different platforms and use cases.

The Problem: When conducting interviews or research at scale, you need to control who accesses your scenarios, limit usage to prevent cost overruns, and make distribution effortless - without complex account management or bulk operations.

The Solution: Generate secure, one-time links that expire after 24 hours (for private scenarios) and can only be used once. Simply click Share → Generate Single-Use Link, configure your options, and send.

Perfect for job candidate screening, user research studies, and any scenario where you need controlled access with accountability.

Add ?maxDuration=900 to any scenario URL (including iframe embeds) to automatically end sessions after the specified time in seconds.

Perfect for timed mock interviews, training with time constraints, and creating consistent evaluation conditions. The AI gracefully wraps up the conversation when time expires.

Customize the AI voice in iframe embeds using a simple query parameter! Override the default scenario voice by adding ?voice=VoiceName to your embed URL.

Available voices: Aoede, Charon, Fenrir, Kore, Puck

Perfect for matching the voice to your brand or audience preferences when embedding scenarios across different contexts.

Create authentic cold calling scenarios with two powerful new features that simulate real-world phone interactions:

• Ring Tone on Call Start - Add a ringing sound when initiating calls, just like dialing a real prospect.

• Instant Hang-up Control - Reduce session end timing to zero, allowing AI agents to hang up immediately when they're not impressed.

Together, these features create hyper-realistic cold calling practice - the phone rings, you deliver your pitch, and if you don't hook them fast, they'll hang up just like a real prospect would.

Build sophisticated multi-stage experiences that seamlessly combine different interaction modes into a single cohesive journey. Perfect for complex sales calls with multiple phases, comprehensive interviews with varied question types, or complete learning paths that blend theory and practice.

Key capabilities:

• Multi-Stage Workflows - Create sales calls that progress through discovery, pitch, and objection handling stages
• Learning Journeys - Combine theory presentations (Google Slides) with hands-on coding practice and MCQ assessments The AI maintains full context across transitions, creating a unified experience that feels natural while providing the depth of multiple specialized scenarios.

Toggle immersive mode on the scenario run page to have the AI avatar fill your entire screen, creating a distraction-free, face-to-face conversation experience. Perfect for deep focus practice sessions where you want to feel like you're truly talking to a real person.

Send supplementary information to embedded scenarios that gets added directly to the session transcript for evaluation. External systems can now inject additional context, observations, or metadata that becomes part of the session record.

Perfect for adding evaluator notes or external data points, or system events that should be included in the final analysis.

Deploy AI agents as silent notetakers to record, transcribe, and evaluate meetings against custom rubrics. The bot joins without video streaming, captures multi-participant conversations, and provides complete transcripts with performance assessments.

Perfect for interviews, training sessions, and group practice scenarios where you need comprehensive evaluation of participant interactions.

Click the magic wand icon in Advanced Scenario Creator to automatically generate avatars based on your scenario's description, voice, and context. No more searching for the perfect image - let AI create it for you!

Scenarios can now be shared and collaborated on with your team members. Add team members with view or edit permissions to build AI roleplay experiences together - perfect for enterprises and teams working collaboratively.

Introducing AI-powered editor for scenario creation that brings together best practices, research tools, and intelligent assistance.

Key capabilities:

• Deep Research - Automatically research topics using Perplexity to enrich scenario content with accurate, up-to-date information
• Smart Templates - Access handcrafted templates embedded with years of scenario writing expertise
• Multi-source Input - Upload PDFs, documents, or provide URLs to automatically incorporate relevant content
• Iterative Refinement - Chat with AI to progressively enhance your scenario through multi-turn conversations

This tool embodies all our knowledge in writing high-quality roleplay scenarios, dramatically reducing time to production while ensuring consistency and quality.

Our AI agents can now use an interactive notepad to write queries, jot down questions, save notes - and read what you write in real-time. This creates a truly interactive learning experience for SQL, programming languages, and technical concepts.

Simply upload PDFs, Word docs, or text files when creating your scenario. Your AI agent will seamlessly reference this knowledge during conversations, providing accurate, context-aware responses based on your specific materials.

Perfect for sales playbooks, interview guides, company handbooks, technical documentation, or any training material you want your AI to reference during conversations.

Our image generation tool now uses Google's cutting-edge Gemini Nano Banana model, bringing powerful image editing capabilities alongside significantly faster generation speeds.

New Features:

• Image Editing - Re-imagine scenario avatar in different scenarios while maintaining character consistency
• Faster Generation - Original image creation is now much quicker Ask the AI to edit existing images: 'Show me in a boardroom setting' or 'Place me in a casual coffee shop environment' - the possibilities are endless!

Your Tough Tongue AI agents can now join Google Meet calls autonomously! Perfect for teams that need help with screening calls, customer data collection, or training sessions.

Two ways to deploy:

• Custom Meeting Mode - Provide the meeting URL and schedule when the bot joins
• Calendar Integration - Connect your calendar and filter by keywords (e.g., 'interview 101')

Getting started: Create your scenario → Go to Meeting Bot → Choose your mode → Let AI handle the conversation

Currently in beta. FREE through August 2025, then available on Premium ($20+) plans.

AI agents can now generate images during conversations, creating immersive visual context that makes practice more realistic.

Use Cases: See the conference room before your presentation, visualize interview settings and professional environments, understand cultural contexts when learning languages, and picture team dynamics for leadership scenarios.

• For Users: Simply ask the AI to show you what something looks like
• For Creators: Add image generation instructions directly in your scenario's AI Instructions to automatically generate relevant visuals at key moments

We're used to consuming content at 1.5x or 2x on YouTube and podcasts - waiting for AI to speak at normal speed can feel painfully slow. Now you can adjust playback speed to match your listening preference! This experimental feature is still a work in progress, but early feedback has been positive.

Finding the right practice scenario just got easier! Based on your feedback, we've introduced:

Tracks - Career-focused learning paths (PM, Engineering, Leadership, Sales). Coming soon: progress tracking and personalized recommendations.

Labels - Instant filters for specific skills. Need Product Analytics or Array Problems? Jump straight to it.

Track your daily practice consistency! Check your streak counter in the Profile menu to see how many consecutive days you've practiced. Miss a day? Your streak resets to zero, but keep going - consistency is key to improving your high-stakes conversation skills.

Our agent can now display and navigate Google Slides based on conversation flow, creating truly interactive learning experiences.

Setup in 3 steps:

1. Get embed URL from Google Slides (File → Share → Publish to web → Embed)
2. Enable tool in scenario settings and paste URL
3. Add slide instructions to AI prompt (e.g., "Show slide 3 for architecture diagram")

Perfect for: System design walkthroughs, product demos, onboarding sessions, technical interviews, and any scenario where visual aids enhance learning.

Making AI conversation practice more accessible! Our new Standard plan offers 100 minutes/month for just $12 - perfect for regular practice sessions. Plus, save big with annual subscriptions:

• Standard Annual at $120/year (save $24)
• Premium Annual at $200/year (save $40)

Share Tough Tongue AI with friends and you'll both get 25 free minutes! Simply copy your unique referral link from the billing page and share it with anyone who could benefit from AI conversation practice. It's our way of saying thank you for spreading the word about high-stakes conversation practice.

Configure scenarios to make the user speak first instead of the AI agent. Control this directly from the Advanced Scenario Creator UI with the 'User speaks first' checkbox. The agent stays silent until you initiate the conversation.

New Galaxy Medium Artifacts model excels at following instructions for generating structured content like cards, MCQ diagrams, and other visual artifacts. This specialized model provides more accurate and consistent output when creating educational materials or interactive content.

Introducing noise cancellation models that intelligently focus on the primary speaker while filtering out background voices and ambient noise. Available in both Galaxy Medium and Galaxy Medium Stable variants, these models ensure clearer conversations in noisy environments, perfect for real-world practice scenarios.

Enhanced avatar customization! In addition to using public URLs, you can now upload avatar images directly from your device. Simply click the upload button in the appearance section to add avatars from your local files - perfect for proprietary images or quick prototyping without hosting.

We've moved Collections to a dedicated Course page to better represent what they truly are - structured learning experiences!

Previously nestled as a tab in the Library, 'Collections' was a technical term that didn't capture the essence of these curated learning paths. It makes it clear that these curated scenario sequences in:

• Product management
• Software development
• Leadership
• Sales coaching

...are actual courses that combine AI roleplay with expert-led sessions.

This clearer structure helps distinguish between:

• Scenario Library → for quick practice scenarios
• Courses → for comprehensive learning journeys

Add {{name}}, {{resume}}, or any variable to your AI Instruction in scenarios. Users get a quick popup to personalize their experience before starting.

Users can simply copy-paste their actual resume, job description, or any relevant document. Perfect for interview prep, sales coaching, and training roleplays where the AI needs your real information to provide meaningful practice.

Two ways to use: (1) Regular scenarios - simply use variable names in your prompt, and users will see a popup to provide the data. (2) Iframe embeds - pass variables as query parameters with 't_' prefix (e.g., ?t_name=John) which maps to {{name}} in your scenario template.

You can now control and edit scenario memory directly from any scenario page! A memory indicator appears when memory is enabled, giving you quick access to modify what AI agents remember you.

Quality of life upgrade: AI agents are now better at conducting longer sessions and less chatty during conversations. The AI will allow you to take longer pauses and better understand when to speak versus when to stay quiet - though as we know, even humans find it tricky to read the room perfectly! We'd love to hear your feedback on this improvement.

Iframe embeds fire an 'onSubmit' event when post-session submission completes. This allows you to redirect users or trigger follow-up actions after required uploads finish.

We've improved how user instructions are displayed in scenarios with a sleek new overlay design. This update is particularly helpful for abstract scenarios like sales coaching and product management, making it easier to understand your objectives at a glance. Try the enhanced experience in our Product Manager collection!

Coding scenarios now support multiple programming languages! Previously limited to Python, you can now practice coding interviews in Python, Java, and C++.

Receive automated analysis reports for sessions conducted with your scenarios! You can configure the email delivery when creating or editing a scenario to get this report delivered to your inbox.

Added a Creator Analytics page that displays all your scenarios with their session counts and related pages in a sortable table.

We've refreshed our homepage with a sleek new design featuring an interactive carousel that showcases pre-created AI role-play scenarios. Quickly browse and discover scenarios to start practicing conversations right away.

We're excited to launch our official blog! Stay up-to-date with product updates, explore technical deep dives, and discover how to get the most out of Tough Tongue AI.

Introducing Scenario Access Tokens - a secure way to embed private scenarios in your applications! This new API endpoint allows scenario creators to generate short-lived access tokens (valid for 1 hour) that enable iframe embedding of private scenarios without exposing your main API key. Key features: (1) Generate temporary access tokens via API for any scenario you own, (2) Tokens expire after 1 hour for enhanced security, (3) Seamless integration with existing iframe embed code - just add ?accessToken=YOUR_TOKEN, (4) Perfect for server-side implementations where you need to control access to private training content. This enables secure, time-limited access to your proprietary scenarios while keeping your API credentials safe.

Introducing Transfer Minutes - a seamless way for coaches, managers, and team leaders to enable AI practice sessions across their entire organization. Instead of managing individual subscriptions, you can now: (1) Buy minutes in bulk (e.g., 2000 minutes), (2) Transfer specific amounts to team members via their email (e.g., 100 minutes each), and (3) Team members simply sign in and start practicing - no payment setup required. This removes all friction for your team while giving you full control over minute allocation and usage. Note: Currently invite-only - email help@getarchieai.com for access.

Two new customization options for iframe embeds: (1) Transparent Background - Set the iframe background to transparent to seamlessly blend with your website's design. (2) Disable Countdown Timer - Control the countdown timer display by adding countdown=false to your embed URL.

Introducing comprehensive memory management for your AI scenarios! You can now view, edit, and manage what AI agents remember about you. Features include: edit 'Super Memory' - the core context that shapes how agents understand and respond to you, manage individual conversation memories that are automatically extracted from past interactions, delete unwanted memories, and access all memories through a dedicated management interface in Settings. This gives you full control over your AI agent's contextual understanding.

Introducing a new feature to control Voice AI agents' capabilities for creating cards, drawing diagrams, uploading PDFs, accessing the code editor, and more! You can now enable or disable specific tools for each scenario and auto-add tool instructions, or manually include specific instructions in your AI Instructions for more precise control.

Introducing Organizations (previously Brand Pages) - create fully customized landing pages for your content with team access control. Features include: custom subdomain URLs (<yourname>.app.toughtongueai.com), branded layouts with logo and cover images, curated scenario and collection selection, social media integration, FAQ sections, public/private visibility controls, and granular team member permissions. Perfect for organizations and trainers who want to share their content with a professional, branded experience while managing team access.

Choose between different AI models for your conversations. Galaxy Medium is the default choice with emotion voice modulation and good non-English language support. Galaxy Medium Stable is recommended for production applications requiring high reliability. Ocean Medium Stable offers an alternative provider optimized for phone conversations. For noise cancellation, use the Galaxy Medium NC or Galaxy Medium Stable NC variants.

Added a new example showing how to create auto-expanding iframes that show audio visualization on first click. This interactive demo includes a code editor for testing implementation, real-time event logging, and a visual guide on how the expansion works.

Added a new 'onClick' event to iframe embeds that fires when users click within the iframe. This can be used to create dynamic experiences like expanding the iframe on interaction, tracking user engagement, or triggering custom behaviors in your application.

Enhancements to our public API include: (1) Session analysis reporting - get comprehensive reports of conversation sessions; (2) Enhanced scenario creation with options to configure auto-submit and auto-analyze behavior; (3) Improved naming flexibility - accounts can now contain scenarios with identical names, eliminating previous unique naming constraints.

You can now customize the background color and title of your iframe embeds to match your brand identity and visual style preferences.

Introduced the ability for users to create and share custom conversation scenarios tailored for specific situations.

Now you can access featured scenarios crafted by us. Library today includes: Kellogg Business School, Google PM Interview, and System Design

First release of Tough Tongue AI, a platform that helps leaders and managers get better at tough conversations.