Convora Documentation
Everything you need to set up, customize, and scale your AI assistant.
Quick Start
Get your AI assistant live on your website in under 10 minutes. No technical expertise required.
1. Create your workspace
Sign up at app.convora.ai and create your workspace. Each workspace is a fully isolated environment with its own knowledge base, branding, and settings.
2. Train your assistant
Give Convora knowledge about your business. You can:
- Crawl your website — enter your URL and Convora reads every page automatically
- Upload documents — PDF, Word, CSV, JSON, HTML, Markdown, or plain text
- Connect your API — for real-time data like inventory, order status, or pricing
3. Embed on your site
Add one code snippet before your closing </body> tag:
That's it. Your AI assistant appears as a floating chat bubble, ready to answer customer questions 24/7.
Embed on Your Site
Convora works with a single script tag. No npm install, no build step, no dependencies.
Customization via attributes
Configuration options
| Option | Description | Default |
|---|---|---|
botName | Display name of your assistant | Convora |
color | Accent color (hex code) | #0891b2 |
theme | light, dark, or auto | auto |
require_email | Require email before chatting | false |
startOpen | Open chat on page load | false |
Configuration
Fine-tune your assistant's behavior from the Client Portal under Widget Settings.
JavaScript API
Push live data into the assistant or trigger actions programmatically:
Website Crawling
Convora reads your entire website and learns from it. It handles JavaScript-rendered pages (React, Vue, Angular) automatically.
- Full site crawl — crawls up to 500 pages from a single URL
- Single page add — add individual pages (new blog posts, product pages)
- Cancel mid-crawl — already-indexed pages are kept
- Smart deduplication — re-crawling won't create duplicate content
Manage crawls from Client Portal → Web Crawls.
File Upload
Upload documents to expand your assistant's knowledge. Supported formats:
| Format | Max Size |
|---|---|
| 25 MB | |
| Word (.docx) | 10 MB |
| CSV, JSON | 10 MB |
| HTML, Markdown, TXT | 10 MB |
Files are indexed within seconds of upload and immediately available for the assistant to reference.
Live Data (API Integration)
Connect your backend API so the assistant can answer questions about real-time data — inventory, order status, pricing, user accounts.
Configure under Widget Settings → API Integration:
- Enter your API base URL
- Add a Bearer token for authentication (optional)
- The assistant automatically calls your API when questions require live data
/api/products?search=blue+hoodie and responds with live inventory data.Brand Voice
Control how your assistant communicates. Configure under Widget Settings → Brand Voice.
Tone presets
| Tone | Style |
|---|---|
| Professional | Warm, concise, business-like |
| Friendly | Casual, approachable, uses emojis |
| Formal | Precise, proper grammar |
| Playful | Fun, energetic, uses humor |
| Technical | Detailed, industry terminology |
| Empathetic | Supportive, understanding |
You can also write custom brand guidelines (e.g., "Always refer to us as 'the team'. Never say 'unfortunately'.") or provide a full system prompt for complete control.
Voice Chat
Visitors can speak to your assistant using their microphone and hear responses read aloud.
- Mic input — tap the microphone button to speak naturally
- AI voice responses — the assistant reads its reply in a natural voice
- Toggle control — visitors can mute/unmute voice at any time
- Word highlighting — text highlights word-by-word as it's read aloud
Works in Chrome, Edge, and Safari. No additional setup required.
Proactive Triggers
Automatically engage visitors based on their behavior instead of waiting for them to click the chat bubble.
| Trigger | When it fires |
|---|---|
| Time on page | After X seconds on a page |
| Scroll depth | When visitor scrolls past X% |
| Exit intent | When mouse moves toward closing the tab |
| URL pattern | On specific pages (e.g., /pricing) |
| Returning visitor | When someone visits for the 2nd+ time |
| Idle timeout | After X seconds of inactivity |
Configure under Client Portal → Engagement → Proactive.
Multi-Language Support
Convora understands and responds in 100+ languages. If your content is in English but a visitor writes in Spanish, Hindi, Arabic, or any other language — the assistant responds in their language using your English content.
No configuration needed. Language detection is automatic.
Client Portal Overview
Your command center for managing everything. Access at app.convora.ai.
| Section | What it does |
|---|---|
| Dashboard | Metrics overview — messages, visitors, leads, knowledge sources |
| Conversations | Every chat thread with full history and visitor context |
| Tickets | Support ticket management with SLA and priority |
| Handoffs | Conversations escalated to human agents |
| Visitors | Everyone who interacted, with profiles and activity |
| Leads | Pipeline view: New → Contacted → Qualified → Converted |
| Analytics | Charts, trends, top questions, country breakdown |
| CSAT | Customer satisfaction scores and feedback |
| Files | Upload and manage knowledge documents |
| Web Crawls | Crawl history, status, pages indexed |
| Automation | If/then rules and visual flow builder |
| Engagement | Proactive triggers, webhooks, macros, A/B tests, knowledge gaps |
| Channels | Omnichannel setup (WhatsApp, Slack, Email, etc.) |
| Widget Settings | Branding, voice, handoff, API integration, embed code |
| API Keys | Manage authentication keys for the widget |
Conversations
Every chat your assistant has is saved and searchable. Conversations are grouped automatically — messages from the same visitor within 30 minutes belong to one conversation.
- Full message history with timestamps
- Star important conversations
- See which visitor sent each message
- Source citations show where each answer came from
- AI-generated conversation summaries
Visitors & Leads
Track every person who interacts with your assistant.
Visitor tracking
Automatically captured: country, browser, pages visited, referrer, session count, first/last seen.
Lead capture
- Optional pre-chat form (name + email required before chatting)
- Pipeline stages: New → Contacted → Qualified → Converted → Lost
- Activity timeline for each lead
- One-click email or call directly from dashboard
- CSV export for CRM import
Analytics
Understand how your assistant is performing with built-in analytics (7, 30, or 90 day views):
- Message volume over time
- Visitor and lead trends
- Top questions asked
- Visitor countries
- Conversation count and average length
- CSAT score trends
Automation Rules
Create if/then rules to automate repetitive tasks. No coding required.
Triggers
New conversation, new ticket, keyword match, visitor action, low confidence response, or scheduled time.
Actions
Assign ticket, send auto-reply, add tag, change status, fire webhook, or escalate to human.
Rules are prioritized and can be toggled on/off individually.
Flow Builder
Design visual conversation flows without writing code.
- Trigger node — starts the flow (keyword, first message, page visit)
- Message node — bot sends a specific message
- Condition node — if/else branching based on user input
- Action node — create ticket, mark lead, send email
- Handoff node — transfer to a human agent
Connect nodes in sequence. Enable/disable flows independently.
Webhooks
Push events to external services in real-time. Connect to Slack, Zapier, Make, n8n, or any URL.
| Event | When it fires |
|---|---|
| new_message | Every new chat message |
| new_lead | Visitor becomes a lead |
| new_ticket | Ticket is created |
| handoff_requested | Human escalation triggered |
| csat_submitted | Visitor rates their experience |
| conversation_ended | 30-min inactivity timeout |
Optional HMAC signature for payload verification. Configure under Engagement → Webhooks.
Ticketing System
Full support ticket management when conversations need follow-up beyond what the AI handles.
- Priority levels — Urgent (1h SLA), High (4h), Medium (8h), Low (24h)
- Status workflow — Open → In Progress → Waiting → Resolved → Closed
- Categories — General, Billing, Technical, Feature Request, Bug Report
- Assignment — Route tickets to specific team members
- Internal notes — Message thread within each ticket
- Metrics — Average resolution time, open count, SLA compliance
Human Handoff
When the AI can't help or a visitor explicitly asks for a human, the system creates a handoff session.
- Visitor says "talk to a human" (or similar)
- Bot acknowledges and creates a handoff session
- Session appears in your dashboard under Handoffs
- Agent claims the session and responds
- Session is marked resolved when complete
Optional email notification when a handoff is requested. Enable under Widget Settings → Handoff.
CSAT & Feedback
Track customer satisfaction with two built-in feedback mechanisms:
- Thumbs up/down — visitors rate individual bot messages (visible on every response)
- 5-star CSAT — satisfaction score after conversations or tickets
View aggregate scores, satisfaction rate (% positive), and individual feedback comments in CSAT dashboard.
Omnichannel
Connect your assistant to multiple channels. One knowledge base powers all of them.
| Channel | How to connect |
|---|---|
| Web Widget | Embed code (included by default) |
| WhatsApp Business API token | |
| Forwarding address | |
| Slack | Slack Bot Token |
| Microsoft Teams | Bot Framework App ID |
| Facebook Messenger | Page access token |
| SMS | Twilio credentials |
Configure under Channels & Integrations.
Marketplace Connectors
Connect your e-commerce platform so the assistant can answer product, order, and inventory questions using live store data.
| Platform | What you need |
|---|---|
| Shopify | Shop domain + Admin API access token |
| WooCommerce | Store URL + Consumer key/secret |
| Amazon | Seller credentials |
| eBay | Developer credentials |
| Magento | Store URL + Integration token |
API Reference
All endpoints require an API key. Get yours from Client Portal → API Keys.
Authentication
Pass your key in the request header:
Send a message
Returns a Server-Sent Events stream with the AI response, token by token.
Rate limits
| Endpoint | Limit |
|---|---|
| Chat | 60 requests/minute |
| Crawl | 5 requests/hour |
| File upload | 30 requests/hour |
| All other | 120 requests/minute |
Security & Privacy
- Data isolation — each workspace is completely isolated. No cross-tenant access.
- Encryption — TLS in transit, encrypted at rest
- No training on your data — your content is never used to train AI models
- API key hashing — keys stored as SHA-256 hashes, never in plaintext
- Content filters — ban words, block IPs, restrict countries
- GDPR-ready — data access, export, and deletion supported
- Data retention — deleted within 30 days of account cancellation
Frequently Asked Questions
How accurate is the assistant?
Convora only answers from your indexed content. If it doesn't find relevant information, it says so honestly rather than guessing. Accuracy depends on how complete your knowledge base is.
Does it replace human support?
For 70-80% of common questions, yes. For complex issues, the assistant captures the lead and you follow up manually — or use the Human Handoff feature for immediate escalation.
How fast are responses?
Typically 3-8 seconds. Responses stream word-by-word so visitors see them building in real-time.
Can multiple team members access the portal?
Yes. The workspace owner can invite team members with appropriate permissions.
What if someone asks something not in my data?
The assistant responds with: "I don't have specific information about that. Feel free to contact us directly." It never fabricates answers.
How do I update the knowledge base?
Re-crawl your site (only new/changed content is added), upload new documents, or add individual pages. Changes are live within seconds.