Mendable

What it is

Mendable is a platform for building and deploying AI-powered chatbots trained on technical documentation and resources. It is designed for businesses, from startups to large enterprises, that need to provide automated technical support and answers to customers and employees. The service allows organizations to connect their existing knowledge bases, documentation, and support channels to create a specialized AI assistant.

Main Features

Data Integration

• Over 20+ managed data connectors for sources like Google Drive, Salesforce, and Zendesk
• Supports integration via a simple online GUI or API
• Continuous data synchronization through CRON jobs and webhooks

Model Customization

• Support for multiple base LLM models, including GPT-3.5-Turbo and GPT-4
• Privacy-first custom private and open source LLM options
• Training through answer correction functionality
• Custom prompt editing capabilities
• Model creativity control settings

Deployment & Components

• Multiple deployment options: React, Vanilla JS, and API integration
• Pre-built components including search bars, chat bubbles, and full CLIs
• Customizable components or ability to build custom implementations

Security & Compliance

• SOC 2 Type II certification
• Single sign-on (SSO) support (SAML, OIDC, OAuth)
• Role-based access control (RBAC) at project and chunk levels
• Bring your own key (BYOK) and bring your own model (BYOM) options
• Rate limiting protection at project and user levels

How it works

Technical Documentation Support

Users connect their technical documentation sources to the platform. The AI chatbot ingests this information and learns to answer technical questions from customers or employees. Users can deploy the chatbot directly into their documentation portals or support channels.

Customer and Employee Assistance

The trained AI model provides instant answers to technical questions, reducing the need for human intervention in support and sales enablement scenarios. When the bot cannot answer a question, it can redirect users to human customer support.

Continuous Model Improvement

Users can correct wrong responses generated by the model, which instantly learns from this feedback. This allows for ongoing improvement of answer accuracy and relevance without requiring technical retraining.

Key Points

• Focuses specifically on technical documentation and resources rather than general knowledge
• Provides source verification to reduce AI hallucinations by grounding answers in actual documentation
• Offers enterprise-grade security features suitable for large organizations
• Includes analytics dashboard for tracking user queries, interactions, and model performance
• Supports both internal (employee-facing) and external (customer-facing) use cases

Additional Details

Pricing

• Free plan available with 500 message credits (reduced to approximately 166 messages if using GPT-4)
• Differentiated pricing for internal vs. external use cases
• Enterprise and custom plans available for high-volume usage and custom requirements
• Open source projects may be eligible for free access

Requirements

• Requires creating an account and obtaining an anonymous key for client-side implementations
• API keys required for server-side implementations
• Supports integration with various documentation platforms and knowledge bases

Availability

• Cloud-based platform with web dashboard
• API access for custom integrations
• Managed ingestion services for data processing
• Currently not open source, though some components and integrations are available