MagicHarbor.ai
Hello
Balance: $0.00
AI Transparency
This page explains how AI systems are used within MagicHarbor.
AI Models
MagicHarbor provides access to multiple third-party AI models, including models from OpenAI, Anthropic, Google, DeepSeek, xAI, and others.
Users can select and switch between models depending on their needs.
Different models may vary in reasoning ability, writing style, coding performance, speed, cost, multimodal capabilities, tool compatibility, and response quality.
MagicHarbor helps users work with these differences in a structured way rather than relying on a single AI model for every task.
AI-Generated Content
Responses generated through MagicHarbor are produced by AI models.
AI-generated outputs may be inaccurate, incomplete, outdated, biased, or unsuitable for a specific purpose.
You should review AI outputs carefully before relying on them, especially for legal, medical, financial, technical, business-critical, or safety-related decisions.
MagicHarbor does not provide professional advice through AI-generated responses.
Multi-Model Interaction
MagicHarbor allows users to interact with different AI models within one workspace.
Users may switch models during a dialogue or compare responses from different providers.
Because each model is built and optimized differently, the same prompt may produce different answers depending on the selected model.
This difference is one of the reasons MagicHarbor provides model switching, analytics, and comparison features.
Head-to-Head Compare
MagicHarbor includes a Head-to-Head Compare feature that allows users to send the same prompt to multiple selected AI models.
This feature is designed to help users compare model behavior under similar conditions.
Head-to-Head Compare may show:
• responses from each selected model
• response time / latency
• token usage
• estimated cost
• clarity, detail, and value indicators
• category winners
• comparison summaries
Head-to-Head results are task-specific. A model that performs well on one prompt may not perform best on another prompt.
Comparison results should be treated as practical guidance, not as absolute or scientific benchmarks.
Model Ranking and Analytics
MagicHarbor may provide analytics and model ranking features to help users understand model usage and performance.
Analytics may include:
• number of messages
• input and output tokens
• estimated cost
• latency
• selected provider and model
• Head-to-Head comparison metrics
• model-level performance indicators
Model Ranking may consider factors such as quality signals, speed, cost efficiency, stability, and usage patterns.
These rankings are intended as guidance only.
Scores and rankings may change over time as models, pricing, provider behavior, evaluation methods, and backend systems evolve.
Use Tools and External Information
MagicHarbor may provide a Use Tools option that allows supported models to access backend-enabled tools, such as online search or external information retrieval.
When Use Tools is enabled, a model may use information retrieved from external sources to produce a more informed response.
This can be useful when a prompt requires current, factual, or external information beyond the model’s built-in knowledge.
Tool-assisted responses may depend on:
• selected model
• provider behavior
• backend configuration
• external service availability
• quality and freshness of retrieved information
• reliability of third-party sources
Tool-assisted responses may still be inaccurate or incomplete and should be reviewed by the user.
Data Processing
User inputs, including prompts, messages, uploaded content, images, and dialogue context, may be sent to selected third-party AI providers for processing.
When Head-to-Head Compare is used, the same prompt may be sent to multiple selected models or providers.
When Use Tools is enabled, parts of a prompt or query may also be processed through backend-enabled tools or external retrieval services.
MagicHarbor may store dialogue history, model responses, usage records, analytics data, comparison results, and technical logs to operate the service, provide user-facing features, calculate usage, and improve reliability.
MagicHarbor does not use your data to train its own AI models.
Third-party AI providers may process data according to their own terms, privacy policies, and technical systems.
AI Services
MagicHarbor provides access to third-party AI services through a unified interface.
We do not control the internal behavior, training data, availability, pricing, or output quality of third-party AI models.
AI providers may update, remove, rename, limit, or modify models over time.
As a result, response quality, speed, cost, and availability may change.
AI Agents
MagicHarbor may support AI agent functionality as part of its broader platform direction.
AI agents may combine models, instructions, tools, knowledge, workflow rules, and business logic to support task-oriented automation.
Agent outputs and actions may still require human review.
Users are responsible for configuring, monitoring, and validating agent behavior before relying on it for important tasks.
User Responsibility
You are responsible for how you use AI outputs.
You should not rely on AI-generated content as the sole basis for legal, medical, financial, technical, business-critical, or safety-related decisions.
You should also avoid submitting passwords, private keys, confidential business information, medical records, financial credentials, personal data of others, or other highly sensitive information unless you are authorized to do so and understand the risks.
Limitations of AI
AI systems do not truly understand information in the same way humans do.
They generate outputs based on patterns, context, model behavior, and available input.
Even when an answer sounds confident, it may still be wrong.
Tool-assisted responses may improve access to current information, but external sources may also contain errors, outdated information, or incomplete context.
Contact
If you have questions about how AI is used on the platform, contact us at:
Last updated: May 2026