crafterai

What is Crafter.ai?

Crafter.ai is an enterprise AI Agents Studio developed by Athics s.r.l. in Italy. It enables customer service, marketing, sales, and HR teams to create AI agents based on large models, RAG knowledge bases, and business processes through a code-free interface.

The platform is capable of handling natural language queries, and it can also be connected to CRM, ERP, email, and ticketing systems via interfaces, Webhooks, and MCP. Human intervention, session analysis, and enterprise governance are the features that set it apart from ordinary web chatbots.

Core functions

• Create enterprise AI agents using a no-code interface
• Connect to mainstream large language models and avoid being locked to a single model.
• Answers are provided using documents, FAQs, manuals, and CRM data via RAG.
• Use the visual Conversation Designer to design controlled processes.
• It can automatically recognize and support more than 40 languages.
• If necessary, the session is handed over to a human agent.
• View real-time KPIs, topic trends, and exportable reports
• Connect to enterprise systems via APIs, Webhooks, and MCP

LLM, RAG, and dialogue design

RAG corporate knowledge base

Crafter.ai allows documents, operation manuals, common questions, and CRM data to be used to enhance generation processes, so that the agents can draw on the company’s own information when providing answers. Nevertheless, version control, access management, and regular updates are still necessary for these materials.

• Import product, policy, and service information in bulk
• Provide the relevant segments to the language model to generate a response.
• Reduce speculation that relies solely on the general knowledge embedded in the model.
• Maintain Q&A and knowledge content by business topic
• Tests revealed unaddressed issues and outdated information.

Visual Conversation Designer

Sales staff can configure dialogue flows in the visual designer without having to write code from scratch. For deterministic tasks such as authentication, data collection, scheduling, and ticket handling, they can make full use of the explicit node constraint model.

• Design welcome messages, menus, and question branches
• Set up data collection and field validation
• Invoke different knowledge or actions based on the user’s intent.
• Control of how to handle unanswerable questions and incorrect inputs
• Configure conditions for transferring to a human agent and rules for session termination.

Multilingual abilities

The official website claims to support over 40 languages and can automatically detect the user’s language, making it suitable for providing support to customers from different countries as well as for serving employees. However, the fact that multiple languages are supported does not mean that each of them handles industry-specific terminology, dialects, and complex processes in exactly the same way.

• Automatically identify the language based on the user’s input.
• Use the same agent to cover multiple markets.
• Reduce the effort required to replicate the entire process for each language.
• Test the key languages and specialized terms separately before going live.

Human in the Loop – manual intervention

When the agent detects a complex issue, user dissatisfaction, or the need for manual authorization, it can elevate the session to a support agent. The takeover panel retains the historical context, preventing the user from having to repeat their explanation from scratch.

• Automatically identify conversations that need to be upgraded.
• Assigned to human staff in the real-time panel
• Show the complete session history to the agent.
• Let humans maintain control in high-risk stages.
• Analyze the reasons for the upgrade and improve the knowledge base.

Analysis and optimization

• Check session volume, satisfaction, and response time.
• Analysis of resolution rate and proportion of manual escalation
• Automatically identify the topics that users ask about most frequently.
• Monitor traffic peaks and channel changes
• Export the report for the business team to review.
• Update the agent continuously based on knowledge gaps.

Enterprise system integration

Crafter.ai offers RESTful interfaces, Webhooks, and MCP support, and it lists various communication channels such as WhatsApp, web chat, Facebook Messenger, CRM, ERP, email, voice bots, and Slack. Some connectors or voice-related functions are available only as part of specific packages or for enterprise projects.

• Embed a chat window on the website
• Users through WhatsApp and social messaging services
• Read customer information from CRM or write leads.
• Connect ERP with internal business processes
• Send notifications via email and Slack
• Use MCP to provide external tools to AI agents.

Customer service scenarios

• Answer questions regarding products, orders, and service policies throughout the day.
• Gather information on the issue background and create a customer service ticket.
• Handle common level-1 support tasks and reduce the workload on manual staff.
• After identifying emotions or complex issues, switch to a human agent.
• Use session analysis to identify frequent failures and knowledge gaps.

Sales and marketing scenarios

• Collect potential customer information through conversations
• Filter and categorize sales leads based on requirements
• Display personalized content to different users
• Connect to CRM and schedule follow-up contacts
• Analyze conversation activities and conversion paths

HR and Employee Services

• Answer questions regarding the onboarding process, policies, and benefits.
• Guide new employees in submitting documents and forms.
• Provide training materials and operational instructions for the positions.
• Reduce the administrative tasks that require repeated explanations from the HR team
• Provide multilingual support for international employees.

How to create the first AI agent

1. Register for a 30-day free trial and select a specific business use case.
2. Enter the agent name, objectives, users, and response boundaries.
3. Import the cleaned documents, FAQs, and corporate information.
4. Configure key processes and exception branches in Conversation Designer.
5. Set system alerts, multiple languages, appearance, and privacy settings.
6. Prepare at least one set of real questions and complete the training or indexing.
7. Check accuracy in the testing environment, as well as handling of manual interventions and sensitive information.
8. After selecting a valid package, publish it and obtain the integration information for the website or channel.

How to launch and maintain it

1. Confirm that the knowledge base has been processed and that at least the necessary FAQs are ready.
2. Check that the package is valid, as well as the message quota and team publishing permissions.
3. Verify the name, functionality, appearance, system messages, and privacy settings.
4. Deploy to the production environment and conduct end-to-end testing on the test page.
5. Deploy the embedded code or connector to the target channel.
6. After modifying the knowledge or processes, update the agent again according to the type of change.
7. Continuously monitor the resolution rate, satisfaction level, manual intervention instances, and failure issues.
8. Expand the capacity before it approaches the upper limit to prevent sudden disruptions to production services.

Which users are it suitable for

• Enterprise teams that require a European data protection approach
• It is hoped that the sales staff will be responsible for maintaining the robots themselves, with a customer service department in place for such purposes.
• Cross-border companies that need multilingual support
• Sales and operations teams that wish to connect AI with CRM and ERP systems
• HR departments that need employee services and onboarding assistance
• Organizations that require human supervision and session analysis

Product advantages

• LLM, RAG, and visual workflow design are integrated into one platform.
• Code-free operations enable business teams to make continuous modifications.
• Supports over 40 languages and automatic language recognition.
• Features manual takeover and a complete session history.
• Provides real-time analysis, topic identification, and report export.
• Supports APIs, Webhooks, MCP, and various communication channels
• A 30-day free trial without a credit card is available.

Usage restrictions

• RAG and large models may still produce errors or omissions.
• Before importing corporate data, it is necessary to handle issues related to permissions, confidentiality, and retention periods.
• Complex CRM, ERP, and voice integration may require technical support.
• Subscription plans limit production capacity based on the number of messages and concurrent tasks.
• The Free plan does not allow the creation of production agents.
• Advanced Analytics and Voicebot are primarily aimed at enterprises.
• The official website displays two pricing options: subscription-based and token-based top-ups.

GDPR and AI governance

Crafter.ai claims to be designed for the European market and to comply with GDPR and the AI Act, emphasizing data control and human oversight. Companies still need to determine, based on their own needs, the role of the data processors, the location of the data, the retention period, any sub-processors involved, and the arrangements for cross-border data transfer.

• Import only the minimum amount of personal data required for business purposes.
• Provide users with clear explanations regarding AI identity and privacy.
• Add manual review for sensitive areas such as health, employment, and finances.
• Restrict the agent’s permissions to read from and write to corporate systems.
• Retain prompts, tool calls, transfers to human agents, and operation logs
• Regularly test for deviations, privilege escalation, prompt injection, and data leakage.

Price of subscription plan

The detailed pricing page offers monthly and annual payment options; the annual payment page shows a discount of about two months. Messages cannot be carried over to the next period, and if there is an excess, additional packages or upgraded plans must be purchased.

Comparison of package features

Chat count and excess rules

The pricing page counts each message sent by a user or each reply from the robot as a separate Chat; therefore, one question-and-answer session may result in two charges. Monthly message quotas cannot be carried over, and additional message packages can be purchased or an upgrade can be opted for if the quota is exceeded.

• When calculating capacity, estimates are based on two-way messages rather than full sessions.
• Long conversations consume more data than single-turn Q&A.
• The number of concurrent tasks determines the ability to serve users simultaneously.
• Conduct stress testing based on peak values and average cycles before going live.

Based on the token top-up method

The FAQ on the Italian-language homepage states that the platform uses a pay-as-you-go model without any subscription requirement; a minimum deposit of 20 euros is required to obtain 100 million tokens. This differs from the billing system described on the detailed subscription page, and it may be related to the new V3 version or different sales channels.

Before making a purchase, it is necessary to access the actual V3 console to check the available billing options for the account, the definitions of tokens, the differences between various models, their validity periods, taxes, and refund policies. The catalog must not list two different prices as part of the same package.

Free trial

• The trial period is 30 days.
• No credit card is required to start the trial.
• It is possible to create and test agents.
• Production releases are still subject to valid plans and configuration requirements.
• Before the trial period ends, it is necessary to export the test results and confirm the subsequent pricing.

Help documentation

Crafter.ai offers a dedicated documentation site that covers processes such as knowledge teaching, training, deployment, updating, and maintenance. To deploy the documentation, it is necessary to have the knowledge base ready, to have a sufficient number of valid questions and answers available, and to confirm the package selection, quota settings, and configuration status.

• The new deployment process usually takes a few minutes to load the resources and models.
• The model needs to be updated after changes to the knowledge base.
• Faster updates to the chat interface can be achieved when only the color or appearance is changed.
• Integrated code, a testing page, and analysis tools are provided after release.
• Reaching the message limit may affect the continuation of services.

GitHub, APIs, and open-source status

As of this verification, no official open-source platform repository that can be confirmed on the Crafter.ai website was found. The Crafter CMS, Craft Agents, and other projects with similar names discovered are not the same product as Athics’ Crafter.ai.

Basic information

Recommendation score

4.2 / 5. Crafter.ai is suitable for companies that place importance on multilingual support, human intervention, compliance with European regulations, and self-service configuration by business users; however, there are clear discrepancies in the pricing structure, and it is necessary to obtain confirmation via the control panel or in writing from sales before making an official purchase.

Frequently Asked Questions

Is Crafter.ai free?

A 30-day free trial is available without the need for a credit card. The Free tier allows testing of agents, but it is not possible to publish production agents.

How much is Crafter.ai?

The detailed pricing page shows that the Basic plan costs 30 euros per month when paid annually, or 36 euros per month when paid monthly; on the home page, token top-ups start at 20 euros, but the actual amount shall be as indicated in the control panel.

Is programming required?

Creating knowledge bases, processes, and basic agents does not require programming. However, complex interfaces, CRM systems, ERP systems, and custom actions may still need developers.

How many languages are supported?

The official website claims to support over 40 languages and can detect the language automatically; before its official launch, industry-specific terms and key processes need to be tested one by one.

Is RAG supported?

Support is available; documents, FAQs, manuals, and corporate data can be used to ensure that responses are based on internal knowledge. RAG cannot completely eliminate errors.

Can I speak to a human?

Yes. The system can identify the need for an upgrade and transfer the session history to a human agent via the control panel.

How is a Chat billed?

The subscription pricing page counts each user message and each robot reply separately; a complete question-and-answer session typically consumes two Chats.

Does it support WhatsApp and CRM?

It supports integration with relevant channels and enterprise systems; the specific connectors, quotas, and implementation costs depend on the solution.

Is it compliant with GDPR?

Although the platform claims to comply with GDPR and the AI Act, buyers still need to verify the location of the data, the sub-processors involved, the retention period, and the contractual responsibilities.

Is Crafter.ai open source?

It is not open source; no official source code repository confirmed on the official website or an open version that can be hosted locally exists.