Frequently Asked Questions

You can sign up using Google, Microsoft, or email (passwordless magic link). No password is required — just choose your preferred sign-in method and follow the prompts.

Start by creating your organization, then upload documents to your content library, create an AI agent, test it in the Playground, and deploy it to your preferred messaging channels.

No. The Trial plan gives you full feature access with limited credits for testing — no credit card required.

Your trial countdown is displayed on the Billing page. You can upgrade to a paid plan at any time to continue using the platform with higher limits.

Use Saba, our built-in AI assistant — open it with Cmd+K (Mac) or Ctrl+K (Windows/Linux), or click the floating icon in the bottom-right corner. You can also visit our documentation at docs.cuneiform.chat or email [email protected].

After signing in, click "Create Organization" and fill in your organization name (required), industry, and company size. You can optionally add a description and support contact details.

Organization name (required), industry (required), and company size (required). Optionally: description, and support contact details (email, phone, website, address) — agents share these when users need human help.

Currently, each user can create and manage one organization. Contact our support team if you need multiple organizations for enterprise use.

Yes. Open Account Management from your profile menu to update your organization name, description, industry, company size, default response language, and support contact details at any time.

When users ask your AI agent for human assistance, the agent shares your support contact details (email, phone, website) so they can reach your team directly.

Go to the Team page and click "Invite Member." Enter their email address and assign a role (Admin or Member). They'll receive an invitation email to join your organization.

Admins have full access to all features including team management, organization settings, and billing. Members can view agents, content, and analytics but cannot modify settings, manage team members, or access billing.

Team size depends on your plan. Trial has limited seats, Starter and Plus support multiple seats, and Enterprise offers unlimited team members.

Yes. Admins can update any team member's role from the Team page.

Go to Agents and click "Create Agent." Fill in the agent name (max 100 characters), optional description, set up its persona, configure LLM settings, and choose knowledge access. Your progress is auto-saved as a draft.

A persona is a system prompt that defines your agent's personality, behavior, and boundaries. You can write one manually or use the Generate button to launch the AI-powered Persona Builder, which guides you through five characteristic categories.

An AI-powered tool with five characteristic categories: Role & Identity, Communication Style, Response Behavior, Operational Boundaries, and Knowledge Integration. Configure each category, then generate your persona. You can save it as a reusable preset for your organization.

Mandatory safety guardrails applied to every agent. They prevent fabrication of facts, require the agent to say "I don't know" when information is unavailable, block impersonation of professionals, correct false premises, and refuse harmful content. Platform Rules cannot be disabled.

Testing: Playground enabled, integrations disabled (default for new agents). Active: Both playground and integrations enabled. Inactive: Everything disabled temporarily. Archived: Hidden from main list, everything disabled.

Yes. Toggle Advanced mode in the LLM Configuration section to choose a specific model and adjust the temperature (0.0-1.0). Default mode uses your organization's tier settings. For factual Q&A, a temperature of 0.3-0.5 is recommended.

When enabled, your agent includes source document references in its responses. This is useful for legal, compliance, or knowledge-sharing use cases where users need to know where information comes from.

It depends on your plan. Trial offers a limited number, Starter and Plus support multiple agents, and Enterprise provides unlimited agents. Check the Agents page for your current limit.

Yes. Cuneiform Chat supports 24 languages including Arabic, French, German, Spanish, Turkish, Chinese, Japanese, and more. Set a default response language at the organization level in Settings, or override it per agent. The agent responds in the configured language while its accuracy and safety rules remain in English.

Supported formats include PDF (text-based and scanned), Word (DOCX, DOC), Plain Text (TXT), Markdown (MD), Rich Text (RTF), PowerPoint (PPTX, PPT), HTML, JSON, and XML. Images, audio, video, archives, and executables are not supported.

On the Content page, drag and drop files or click the upload area to browse. You can also import from Google Drive, Dropbox, OneDrive, or Box, or switch to the Website tab to crawl public websites. You can assign tags before uploading, and batch upload is supported.

Most documents complete within a few minutes. Larger or more complex files may take longer. You can track progress on the Content page — each document shows its current processing status.

Uploading: File is being uploaded. Queued: Waiting to be processed. Processing: Being analyzed and indexed. Ready: Available for agents to use. Retrying: Processing is being retried. Failed: An error occurred — you can try reprocessing from the document menu.

Folders provide hierarchical organization — each document belongs to one folder. Tags allow cross-cutting categorization — each document can have unlimited tags. Use folders for broad categories (departments, projects) and tags for topics or themes.

Yes, through Knowledge Access Control. Exclude Mode (default) gives the agent access to all documents except those you select. Include Mode restricts the agent to only the folders, tags, or individual documents you select. Changes take effect immediately.

Delete the old version and upload the updated file. The system detects duplicate content automatically. You can also reprocess an existing document from its menu if the original file hasn't changed.

Click the document menu and select "Reprocess." If it continues to fail, verify the file format is supported, the file isn't corrupted, and it's within size limits. Scanned PDFs may have lower quality extraction due to OCR limitations.

Yes. Switch to the Website tab on the Content page, enter a URL, and click Crawl. The crawler visits pages up to 2 levels deep and extracts content as Markdown. You can review and edit each page before processing it into your content library.

The maximum number of pages per crawl depends on your plan. You can crawl one website at a time. Only publicly accessible pages are supported — login-protected content cannot be crawled.

Google Drive, Dropbox, OneDrive, and Box. Each provider has a dedicated file picker accessible from the Content page.

On the Content page, click the icon for your provider, sign in to authorize read access, then select the files you want to import. Selected files are copied into your content library automatically.

No. Cloud imports are one-time copies. If the original file changes in your cloud storage, you'll need to re-import it to update the content in your library.

No. We only request read access to the files you explicitly select. Your original cloud files remain unchanged. You can revoke access at any time through your cloud provider's settings.

Yes. Google Workspace files (Docs, Slides, Sheets) are automatically converted to text during import. Note that formatting, images, and charts are not preserved — only the text content is extracted.

Click "Test" on the agent card or the "Test Agent" button when editing. The Playground opens as a side panel where you can chat with your agent using your actual content. New agents start in Testing status with the Playground enabled.

Test basic Q&A accuracy with your documents, edge cases (unanswered topics, ambiguous questions), multi-turn conversations, topic changes, and tone consistency. Evaluate each response for accuracy, relevance, completeness, and appropriate length.

No. The Playground is only available for agents in Testing or Active status. Change your agent's status to Testing to re-enable playground access.

Your agent is ready when it answers common questions accurately, handles edge cases gracefully, maintains the right tone, respects defined boundaries, and multi-turn conversations flow naturally. Then change its status to Active to enable channel integrations.

Telegram, Discord, Slack, WhatsApp, Google Chat, Meta (Facebook Messenger), and your own website via an embeddable chat widget. All channels are available on every plan. Telegram, Slack, and Google Chat each have their own dedicated FAQ category below.

Yes. Deploy a single agent to multiple channels simultaneously. It uses the same content, provides consistent responses, and all conversations are tracked in unified analytics with centralized history.

At least one agent created, documents uploaded to your content library, and the agent tested in the Playground. Each channel also has specific requirements — for example, Telegram needs a BotFather token, Discord needs app credentials, and Google Chat requires a verified domain.

Create an application in the Discord Developer Portal, add a bot, get the credentials (Application ID, Public Key, Bot Token). Enter these in your agent's Integrations tab, set the webhook URL, then add the bot to your server.

Go to your agent's Integrations tab and select WhatsApp. Use the embedded signup flow to connect your WhatsApp Business Account, verify your phone number via SMS or voice call, and set up Two-Step Verification with a PIN. Messages are subject to WhatsApp's 24-hour messaging window.

Go to your agent's Integrations tab and select Meta. Click Connect with Facebook, authorize your Facebook Page, and grant the required permissions. Your agent will respond with typing indicators and automatically split long messages (2,000 character limit per message). Note: one agent per Page, and Messenger only (not Page comments).

Go to the Integrations page and find the Website Widget section. Click "Create Widget" to configure your widget's appearance, select the agent that will respond, and optionally restrict which domains can embed it. Once created, copy the embed code (a single script tag) and paste it into your website's HTML before the closing '<'/body'>' tag.

Yes. Channel integrations are only enabled when the agent's status is Active. Agents in Testing, Inactive, or Archived status cannot receive messages from external channels. The website widget requires the assigned agent to be Active as well.

Each channel has a step-by-step guide in our documentation at docs.cuneiform.chat/integrations, including screenshots, troubleshooting tips, and security best practices.

Create a bot via @BotFather on Telegram to get a token. Then go to your agent's Integrations tab, select Telegram, paste the token, and connect. You can customize the bot's name, description, and avatar through BotFather. Once connected, the bot starts replying to DMs immediately.

Live streaming replies that land in milliseconds and grow in place as the agent thinks, voice notes auto-transcribed with language auto-detection, rich markdown formatting (bold, italics, code, lists, links), one-tap emoji reactions classified as positive/negative/neutral feedback, and instant /start, /help, /reset, /new slash commands handled with zero AI cost.

Yes. Voice notes are transcribed with studio-grade accuracy and the language is auto-detected from the conversation, so short clips in any language stay accurate. Available on every plan, including the free trial.

Every t.me/your-bot?start=your-code link can carry a custom welcome message and up to three suggested replies. Click-throughs, unique users, and conversation conversion are tracked per code, so you can run different campaigns from the same bot and measure each.

A multipage, in-app catalog that opens inside Telegram when customers tap the chat-menu button next to the message composer — no website redirect, no app switch, no auth dance. Every Telegram bot you connect ships with a Storefront. The chat-menu button is auto-wired, so no BotFather setup is needed.

12 block types — product cards, galleries, videos, themed pages, testimonials, contact cards, FAQs, and more. Six themes, drag-and-drop composer with live preview. Pages are organized in a menu that customers can navigate inside the Mini App.

Yes. Attach knowledge-base documents to a product block and customers can tap and ask "does this come in size M?" with answers scoped only to that product's documents. The chat surface is built into the product page in the Mini App.

Yes. Each product block can carry an order CTA in inquiry or cart mode with configurable form fields. Orders land in your inbox as a structured handoff and the customer gets an instant "got your order" DM back from the bot.

Yes — including the free trial. The chat-menu button is auto-wired the moment you connect a Telegram bot, and a built-in Help Center is pinned at the end of the page menu for ad-hoc FAQs even before you publish any product pages.

Telegram Business is a Telegram Premium feature that lets a bot answer DMs from a user's personal account. Founders, creators, and small teams who get DM'd on their personal Telegram all day can have an agent reply in their voice — same agent, same memory their customers know. Available on every paid tier, including the free trial.

On the Premium user's phone: open Settings → Telegram Business → ChatBots and paste your bot link. Then DM /start your-code to bind the account. The bot answers DMs in your name until you reply manually — once you do, the bot pauses on that chat and resumes only after the conversation goes quiet. A connection-level kill switch in the admin panel disables auto-reply across the whole Premium account.

A dedicated Telegram bot for admins, separate from your customer-facing bot. Link your admin account with a one-time code, then DM Saba for quick text replies, or open the in-Telegram Mini App for a richer chat surface — conversation list, tool-call detail, voice input, markdown export. Saba's replies come with tappable buttons that jump straight to the right panel page, so a question turns into action in one tap.

Go to your agent's Integrations tab and click Connect for Slack. Complete the OAuth flow to authorize your workspace — no webhook configuration, no setup files. Once authorized, the bot is ready to respond in seconds.

Yes. When you @mention the bot in a channel thread, it reads the full conversation history before responding. No copy-pasting context — the bot already knows what the team discussed.

Yes. Each channel can be assigned its own AI agent. Sales gets the product expert, Support gets the troubleshooter — one Slack app, many specialists. Use /set-agent inside a channel to assign one.

/ask for quick questions, /help to see what's available, /set-agent to assign an agent to the current channel, /current-agent to inspect the assignment, and /reset-agent to revert. Full control without leaving Slack.

Yes. The bot answers in channels, threads, direct messages, and group DMs — wherever your team works. Same agent, same memory, every Slack surface.

No. OAuth handles everything — no webhook URL, no signing secrets, no environment variables to manage. The setup is one click from your agent's Integrations tab.

Install it from the Google Workspace Marketplace like any other Workspace app — standard admin deployment, no custom webhooks, no code. The integration setup page in your admin panel links directly to the Marketplace listing.

Domain verification proves you own your organization's domain. It's a one-time setup — add the DNS TXT record provided during integration to your domain's DNS settings. Once verified, only members of your verified domain can interact with your bot.

Most providers propagate within a few hours, but Google allows up to 48 hours. Once the TXT record is verified, it applies to all future Google Chat integrations for your organization — you only do this once per domain.

A Space manager runs /connect once in the Space. After that, the bot gains full conversation lookback for that Space and reads what the team discussed before responding — no more repeating context.

Yes. Different Spaces can be assigned different AI agents — HR gets the policy expert, IT gets the troubleshooter, Sales gets the product specialist. Use /set-agent inside a Space to assign one.

/help to see what's available, /connect to authorize thread history access in a Space, /set-agent to assign an agent, /current-agent to inspect the assignment, and /reset-agent to revert. Full control without leaving the Space.

No. Domain verification ensures only members of your verified domain can access your bot — enterprise-grade isolation, no unauthorized external access.

The dashboard shows total queries, conversations, unique users, and credits used. It includes query volume charts over time, channel and agent distribution breakdowns, and rankings of your top agents and channels by usage.

Yes. Filter by specific agent, channel (Telegram, Discord, Slack, WhatsApp, Google Chat, Meta, Widget, Playground), and date range. Top agents and top channels are ranked by queries, conversations, and credits consumed.

Current billing cycle, last 7 days, last 14 days, last 30 days, or last 90 days. Data can be grouped by hourly, daily, weekly, or monthly granularity.

Identify which agents handle the most queries, find high-volume time periods, review conversation transcripts to spot gaps in your content, and use AI-powered insights to understand user intents and common topics.

Yes. The Conversations section shows individual chat sessions with full message history, timestamps, channel, agent, credits spent, and token usage. You can generate AI-powered insights (key points, user intents, suggested actions) and export conversations as JSON.

Trial (free, limited credits, no credit card), Starter (core features, standard credits), Plus (generous credits, priority support, multiple seats), and Enterprise (custom credits, unlimited agents and seats, dedicated support). All plans include all channel integrations, website widget, and full analytics.

Platform Credits are consumed by AI queries (when agents respond), document processing (when files are uploaded and indexed), and embeddings generation. Your monthly allocation depends on your plan and resets each billing period.

Yes. Upgrades take effect immediately with prorated charges. Downgrades are scheduled for the end of your current billing period. Once a downgrade is scheduled, no further plan changes can be made until the new billing period begins. Verify your current usage fits within the new plan's limits before downgrading.

You'll see a warning at approximately 80% usage. At 100%, a "Limit Reached" banner appears. Service is paused until your next billing period unless you enable Extra Usage or upgrade your plan.

A pay-as-you-go option for paid plans (Starter, Plus, Enterprise) that lets you continue using the platform after your monthly quota is exhausted. Enable it from the Billing page with a minimum $10 top-up, and set a spending cap between $10-$500 per month.

No. Your monthly credit allocation resets at the start of each billing period. Extra Usage balance, however, does carry over between periods.

Go to Billing and click "Cancel Subscription." Access continues until the end of your current billing period. Your agents will stop responding after that, but documents are retained for a limited time so you can reactivate later.

Go to Billing and look in the History section. You can view or download PDF invoices showing the billing period, plan charges, any additional charges, and payment status.

Saba is our built-in AI platform assistant, named after Nisaba — the Sumerian goddess of writing and scribal arts. Saba helps you navigate the platform, answers questions about features, and can access your organization's data to provide specific information about your agents, documents, and usage.

Click the floating Saba icon in the bottom-right corner, or use Cmd+K (Mac) / Ctrl+K (Windows/Linux). Press Cmd+N / Ctrl+N to start a new conversation, and Escape to close.

Platform feature guidance, your organization data (agent counts, document lists, integration status, billing info), and step-by-step workflows. Use slash commands like /agents, /usage, /docs, /help, /new, and /conversations for quick access to common tasks.

Yes. Fast mode gives quick responses for simple questions. Reasoning mode provides more thorough analysis for complex topics. Toggle between them in the Saba panel — your preference is saved across sessions.

Yes. Click the Globe icon in Saba's header to choose from 24 supported languages. Saba defaults to your organization's configured language (set in Organization Settings), but you can override it per session. Your selection is saved in your browser and remembered across sessions.

Yes. All data is encrypted in transit via HTTPS. We use strict tenant isolation (each organization's data is completely separated), role-based access control, and encrypted credential storage. Your documents and conversations are never shared with third parties or used to train AI models.

Models from OpenAI, Anthropic, and Google Gemini. The available models depend on your plan. You can choose the model that best fits your use case, budget, and performance requirements in each agent's Advanced LLM configuration.

Yes. Deploy agents to seven messaging channels (Telegram, Discord, Slack, WhatsApp, Google Chat, Meta, and Website Widget). Import documents from four cloud storage providers (Google Drive, Dropbox, OneDrive, Box). Enterprise customers can access additional integration support.

Your agents stop responding, but documents and conversation data are retained for a limited time. If you reactivate your subscription, your existing data becomes available again. Contact support for specific data retention questions.