
Wit.ai
Customer messages come in many forms but mean similar things. Your AI agent uses Wit.ai to parse text into structured intents, entities, and traits. When someone types 'cancel my 3pm appointment tomorrow,' Wit.ai extracts the intent, date, and time. Your agent acts on structured understanding, not keyword guessing.




Your AI agent uses Wit.ai's NLP engine to decompose customer messages into intents, entities, and traits, turning unstructured text into actionable data.
Wit.ai
From booking requests to support triage, see how businesses use AI agents with Wit.ai to understand what customers mean and act on it automatically.
A customer messages 'I want to return the shoes I bought last Tuesday.' Your AI Agent sends the text to Wit.ai, which extracts the 'return_product' intent, 'shoes' as the product entity, and 'last Tuesday' as the date entity. The agent routes the request to the returns workflow with all details pre-filled. The customer starts their return immediately without repeating any information to a human agent.
Your business operates in 10 countries. A French customer types 'Je voudrais annuler mon rendez-vous.' Your AI Agent routes the message through the French Wit.ai app, extracting the cancellation intent and appointment entity. The agent processes the cancellation using the same backend workflow as English requests. One AI agent serves every language with consistent accuracy.
A customer starts calmly but their frustration builds across messages. Your AI Agent analyzes each message with Wit.ai's trait detection, tracking sentiment shifts from neutral to negative. When frustration crosses a threshold, the agent automatically escalates to a human agent with full conversation context. Angry customers reach humans faster. Satisfied customers stay with the efficient AI.

Wit.ai
FAQs
The agent sends the customer's text to Wit.ai's GET /message endpoint. Wit.ai returns a structured response with detected intents (what the user wants), entities (specific data like dates, locations, amounts), and traits (message characteristics like sentiment). The agent uses these structured results to determine the appropriate action.
The Server Access Token is a per-app authentication credential found in your Wit.ai app settings. Each Wit.ai app has its own token. This token authenticates API requests from Tars and determines which NLP model processes the messages. It is stored securely and used only during active conversations.
The agent can create new Wit.ai apps with specified names, language locales, and privacy settings. It can also create traits with custom values for classification. Training with specific utterances should be done through the Wit.ai dashboard or dedicated training workflows for best results.
No. Tars sends messages to Wit.ai's API in real time and receives structured intent/entity results. The raw message text is not stored or logged by Tars after the NLP analysis completes. Wit.ai may process messages according to their own data policy as part of Meta's platform.
Wit.ai's built-in NLP supports over 10 languages including English, French, German, Spanish, Chinese, Dutch, Italian, Polish, Portuguese, Romanian, and Vietnamese. Each Wit.ai app is configured for a specific language locale, and you can create separate apps for different languages.
Building a custom Wit.ai bot requires significant development effort for hosting, conversation management, and API integration. Tars provides the complete conversation infrastructure, deployment across channels, and integration with 600+ tools. Wit.ai adds deep NLP understanding while Tars handles everything else.
Yes. Wit.ai extracts all recognized entities from a message simultaneously. For example, 'Book a table for 4 at 7pm on Friday in Manhattan' yields entities for number (4), time (7pm), date (Friday), and location (Manhattan). The agent receives all extracted data in a single API response.
You control when Wit.ai is invoked. The agent can be configured to use Wit.ai for complex intent detection scenarios where understanding nuanced customer requests matters most. For simpler interactions, the agent's built-in language understanding handles the conversation without an additional API call.
Don't limit your AI Agent to basic conversations. Watch how to configure and add powerful tools making your agent smarter and more functional.

Privacy & Security
At Tars, we take privacy and security very seriously. We are compliant with GDPR, ISO, SOC 2, and HIPAA.