Supported AI Providers
AgentQ runs on the AI provider configured for your deployment: the Qualytics-managed provider (Beta) or an external provider connected with your organization's credentials. This page lists every provider you can connect, explains what the Beta badge on the Qualytics provider means, and covers which providers support file attachments in chat. To connect or change a provider, see Add Integration and Update Integration.
How the Configuration Applies
The AI provider configuration is a single, deployment-wide setting rather than a per-user choice. A Manager or Admin selects the provider and completes any credentials it requires once, and AgentQ then becomes available to users with the Member role or higher. Only one provider configuration is active at a time.
AgentQ combines the Business Context with each user's conversation context to make responses and suggestions more relevant to your organization's domain. Do not put secrets or anything your configured AI provider should not process into it.
Provider List
The table below covers which providers you can connect. For the models each one currently offers, open the Model dropdown under Custom setup on the AgentQ Settings page after selecting the provider: that list comes from the platform itself and stays current as providers release and retire models. Where a provider accepts any model, you can also type a model name that is not in the list.
| Provider | Models | File Uploads |
|---|---|---|
| Qualytics1 | Managed by Qualytics, with no model to select | |
| Alibaba Cloud | qwen-maxqwen-plusqwen-turbo |
|
| Amazon Bedrock2 | Region-specific model IDs | |
| Anthropic | claude-opus-5claude-sonnet-5claude-haiku-4-5 |
|
| Azure OpenAI | OpenAI models via Azure deployments | |
| Cerebras | gpt-oss-120bqwen-3-235b-a22b-instruct-2507zai-glm-4.7 |
|
| Cohere | command-r-pluscommand-r |
|
| DeepSeek | deepseek-chatdeepseek-reasoner |
|
| Fireworks | Any Fireworks-hosted model | |
| GitHub Models | GitHub-hosted models | |
| Google Gemini | gemini-3.6-flashgemini-3.1-pro-previewgemini-3.5-flash-lite |
|
| Google Vertex AI | gemini-3.6-flashgemini-3.1-pro-preview |
|
| Groq | openai/gpt-oss-120b | |
| Heroku | Claude, GPT-OSS, Nova, and other models via Heroku Managed Inference | |
| Hugging Face | Inference endpoint models | |
| LiteLLM3 | Any model via a LiteLLM proxy | |
| Mistral | mistral-large-latestmistral-small-latest |
|
| Moonshot AI | kimi-k3kimi-latestkimi-thinking-preview |
|
| Nebius AI | Qwen/Qwen3-235B-A22Bmeta-llama/Llama-4-Maverick-17B-128E-Instruct |
|
| Ollama4 | Any locally hosted model | |
| OpenAI | gpt-5.4gpt-5.4-minigpt-5.4-nano |
|
| OpenRouter | Any model via OpenRouter | |
| OVHcloud AI5 | Any OVHcloud AI endpoint model | |
| Perplexity | sonar-prosonar |
|
| SambaNova | Llama-4-Maverick-17B-128E-InstructQwen3-235B-A22B |
|
| Snowflake Cortex6 | claude-opus-4-8claude-opus-4-7claude-sonnet-4-6 |
|
| Together AI | Any Together AI model | |
| Vercel AI7 | Any model via the Vercel AI gateway | |
| xAI | grok-4.5 | |
| Zai | GLM 5 generation models |
Notes
1 Qualytics. Runs on the Qualytics AI gateway, which chooses and updates the model for you. Marked Beta, see What Beta Means.
2 Amazon Bedrock. Model IDs differ by AWS region, and the provider needs a region plus an authentication method. See Amazon Bedrock Authentication.
3 LiteLLM. Needs a Base URL pointing at your proxy.
4 Ollama. Needs a Base URL pointing at your Ollama host.
5 OVHcloud AI. The Model field is free text, so you type the model name rather than picking it from a list.
6 Snowflake Cortex. Every model needs a Snowflake Account Identifier, in orgname-accountname form.
7 Vercel AI. The Model field is free text, so you type the model name rather than picking it from a list.
Note
The Qualytics provider does not require your organization to supply an API key. For an external provider, your organization supplies the credentials and is responsible for that provider's usage and commercial terms.
When the Qualytics Provider Is Not Listed
The Qualytics provider appears only on deployments that carry a Qualytics-issued deployment identifier. Where that is missing, the provider is simply absent from the list: there is no greyed-out entry and nothing on the form explaining why. If you expect to see it and do not, contact Qualytics support, or configure an external provider instead. Selecting it through the API on such a deployment is refused with a message saying the same thing.
What Beta Means
The Beta badge marks this provider where it turns up: on the Easy setup card on the AgentQ Settings page and in the disconnect confirmation dialog. Beta here means capacity is still being expanded while feedback is gathered, so its behavior may change. Everything else about it works the way the other providers do: you connect it, change it, and disconnect it from the same place.
File Uploads in Chat
The File Uploads column above says which providers take attachments. On those, the chat input shows Attach file for sending PDFs, Word, Excel, CSV, TSV, JSON, XML, plain text, and Markdown files. On Amazon Bedrock the button appears for every model, but only the Claude family can actually read an attachment, and another model declines it with a message. Providers without the column marked can still receive document content through the Paste Large Content flow. See Attach a File for limits and supported formats.
The Qualytics provider reads a narrower set of formats
The Qualytics provider accepts PDF, Word .docx, Excel .xlsx, and text-based files (CSV, TSV, plain text, JSON, XML, Markdown). It cannot read the legacy .doc and .xls formats, PowerPoint, or images, so the file picker leaves those out while this provider is selected. Dragging one onto the chat instead of picking it is turned away right away, with a notice listing the formats that do work. Save the file as .docx or .xlsx, export it as CSV or PDF, or paste its content as text.
A PDF has to be text-searchable. A scanned or image-only PDF is still accepted, but no text can be read from it, so AgentQ answers that it found no readable content rather than returning an error. Very long documents are truncated.
See Also
-
Amazon Bedrock Authentication
The three ways to authenticate to Bedrock, and what an IAM role setup expects.
-
Conversations, Responses & Context
How to write prompts, read AgentQ responses, and work with context-aware chats.
-
AgentQ Limits
Rate limits, token usage, timeouts, SQL constraints, and scope constraints.
-
Examples
Worked scenarios showing what you ask AgentQ, what it does, and the shape of the answer.
-
Best Practices
Prompt design, cost management, guardrail behavior, rate limits, and async operation patterns.
-
Permissions
The user roles behind chatting with AgentQ, configuring it, and reading the audit.
-
How It Works
Where AgentQ appears, how a turn runs, and what it is allowed to see.
-
The Chat Interface
Every control in the full-page and floating chat, and what the input accepts.
-
MCP
What the Model Context Protocol is, how it works, and why it matters.
-
AgentQ in Action
How Qualytics implements MCP, with its endpoint, tools, and tool step labels.
-
Tool Catalog
Every tool AgentQ can call, what each one does, and what it shares with the model.
-
Semantic Reporting
The structured query interface AgentQ uses to count, filter, and inspect platform resources.
-
Access Controls
How an administrator chooses what chat may share with the model, and what each level unlocks.
-
AgentQ Audit
What each audit entry holds, how cost estimates work, and what the period summary reports.