Visible steps
The answer shows its progress: validation, context analysis, documents found and tools running.
Good morning!
Good afternoon!
Good evening!
Hello!
Let's get to work
Libraries
Files
Files
Files
Files
Connectors
Shared
Public
My team
Library
Chat commands
GPT-5.5
OpenAI
Complex reasoning and coding
GPT-5.4 Mini
OpenAI
Low latency for high-volume tasks
Sonnet 5
Anthropic
High quality with better speed
Fable 5
Anthropic
Anthropic's most capable model
Gemini 3.1 Pro
Advanced multimodal reasoning
Preview model — may have limitations
Gemini 3.5 Flash
Multimodal speed
Grok 4.5
xAI
Reasoning and coding with a long context
Preview model — may have limitations
Qwen3.7 Max
Alibaba Qwen
Alibaba's flagship for agentic tasks
Seed 2.0 Pro
ByteDance Volcengine
Multimodal with deep reasoning
DeepSeek V4 Pro
DeepSeek
Reasoning with a one-million-token context
Kimi K2.6
Moonshot AI
Multimodal for long coding tasks
GLM-5.2
Z.ai (Zhipu)
Z.ai's flagship with a long context
How should actions be approved?
Anonymization active: sensitive data is replaced with placeholders before it is sent to the AI provider, and restored in the response.
NiuCore may make mistakes. Verify important information.
You can continue with
Sending the message opens the NiuCore sign-up. The text you type is not sent or stored.
Models
The employee chooses among the models enabled for their area or lets Auto mode pick the right one for each task. For every model, the selector shows roughly how many messages the balance allows and its capabilities, such as reasoning or image generation.
Action approval
When an answer requires acting on another tool — sending an email, creating an event, editing a document — NiuCore asks for approval before running it. The employee can change that mode so NiuCore approves on their behalf, or ask it to present an execution plan first to review and approve.
More in every conversation
The answer shows its progress: validation, context analysis, documents found and tools running.
If a provider is unavailable, Auto mode continues with another enabled model and says so in the conversation.
Up to 5 files per message, up to 5 MB each, which NiuCore reads as part of the query.
Generated images and chat attachments are saved to Google Drive, OneDrive or a library, or sent via Gmail.
An indicator shows how much context the conversation uses, which is compacted to free space as it nears the limit.
Sensitive data is replaced with placeholders before it reaches the model and restored in the answer.
Before you subscribe