# AI Platform FAQ

You need an active PHOENIQS Model Service API key before purchasing a PHOENIQS Chat plan. Once subscribed, access depends on your plan: the free Evaluation plan or the Business plan with a dedicated single-tenant environment. If your company already has a paid subscription, your administrator adds you as a user and shares your access URL (for example, chat.your-company.ch). Paid plans use enterprise SSO — you sign in through your company identity provider and never type a password directly into PHOENIQS Chat.

No. Your data is never used to train the underlying models, and model vendors never see your data. Interactions run on Swiss infrastructure inside confidential-computing enclaves and are protected by your own encryption keys. Session history is kept locally, and you retain full ownership and control of your data. This applies across both Evaluation and Business plans.

The Evaluation plan is free and lets you try PHOENIQS Chat on Swiss-hosted infrastructure with SSO, RAG, MCP integrations, web search, and 1 million tokens per month. The Business plan (CHF 599/month excl. VAT) adds a dedicated single-tenant private instance per company, the full catalogue of Swiss-hosted models, Microsoft 365 and Atlassian integrations, pre-built reusable assistants, custom branding, enterprise SSO via Keycloak, 1 GB of RAG storage, and a CHF 80 universal token credit pool — both storage and credits can be increased on demand. Business also includes a basic SLA.

A regular conversation is a one-shot exchange: you ask, the model answers, and context from that session does not carry forward unless you provide it again. A custom AI assistant (agent) is a reusable, no-code configuration you build once in the Agent Builder — with a name, instructions, model, knowledge files, and connected tools. The assistant remembers its role and sources every time you open it, so repeatable tasks like policy Q&A, document review, or training programme design do not require re-explaining context or re-attaching files. You can share assistants and conversations with colleagues, and connect them to company tools such as Microsoft 365 and Atlassian via MCP.

The most common causes include:

• Excessive context passed through MCP (Model Context Protocol)
• Tool-calling loops, where each tool result is injected back into the prompt
• Long conversation history that is continuously appended without truncation
• Large prompts or system instructions
• Token estimation mismatches due to tokenizer differences or fields added late in the request pipeline

When combined, these factors can quickly exceed the model's context window.

A 400 Bad Request error related to tokens typically occurs when the total estimated input tokens exceed the model's context window.

Phoeniqs' AI gateway (LiteLLM) performs a protective pre-check before sending the request to the model. If the request would exceed the model's allowed context window, LiteLLM rejects the request early to avoid unnecessary compute usage. This situation can result in zero or negative tokens remaining for the model's response, which triggers the error.