Frequently Asked Questions

General

What is the Chat App?

The Chat App is FlowState's multi-provider LLM chat interface. It enables you to have real-time streaming conversations with AI models from Anthropic (Claude), OpenAI (GPT), and local models via LM Studio. All conversations are persistently stored with organization and workspace scoping.

Who should use this app?

The Chat App is designed for:

User Type	Primary Use Cases
Developers	Code assistance, debugging, API exploration, tool execution via MCP
Knowledge Workers	Document drafting, research, Q&A with persistent conversation history
Teams	Shared workspace conversations with organizational context
Power Users	Local model integration via LM Studio, custom system prompts

How is this different from using ChatGPT or Claude directly?

The Chat App offers several advantages:

Multi-Provider Support: Switch between Anthropic, OpenAI, and local models without changing apps
Persistent History: Conversations are stored in your FlowState database, not just browser sessions
Organization Scoping: Conversations are organized by workspace and team
MCP Tool Integration: AI can execute tools via FlowState's MCP Gateway
Local-First: Works offline with RxDB's local-first architecture

Features

How do I switch between AI providers?

To switch providers:

Open any conversation
Click the model dropdown in the header
Select your desired provider (Anthropic, OpenAI, or LM Studio)
Choose a specific model from that provider

The new model will be used for subsequent messages in that conversation.

Can I use custom system prompts?

Yes, each conversation can have a custom system prompt. System prompts help guide the AI's behavior and tone. The system prompt indicator in the header shows when a custom prompt is active.

What models are available?

Anthropic Claude:

Claude Sonnet 4.5 (balanced performance)
Claude Opus 4.1 (most capable)
Claude Haiku 4.5 (fastest)
Previous generation models (4.0, 4)

OpenAI GPT:

GPT-4o (latest flagship)
GPT-4 Turbo (fast and capable)
GPT-3.5 Turbo (fast and economical)

LM Studio:

Any model you have loaded locally

How do MCP tools work?

When the FlowState MCP Gateway is configured:

Available tools are automatically discovered when the app loads
The AI can invoke these tools during conversations
Tool calls appear with expandable sections showing arguments and results
Status indicators show pending, success, or error states

Tools enable the AI to perform actions like querying databases, creating records, or interacting with external services.

Is markdown supported in messages?

Yes, the Chat App provides full markdown support including:

Headers, bold, italic, strikethrough
Code blocks with syntax highlighting for many languages
Lists (ordered and unordered)
Tables
Links and images
GitHub Flavored Markdown (GFM) extensions

Account and Access

How do I get access to this app?

The Chat App is included with your FlowState workspace. Ensure you:

Have an active FlowState account
Are a member of a workspace
Have the necessary permissions to access apps

What permissions do I need?

You need the following permissions:

Database access: To read and write conversations and messages
API access: To communicate with LLM providers (via configured API keys)

Are my conversations private?

Conversations are scoped to your organization and workspace. This means:

Only members of your organization can see organization conversations
Workspace conversations are visible to workspace members
API keys are managed at the organization level, not shared in conversations

Technical

Which browsers are supported?

We support the latest versions of:

Chrome (recommended)
Firefox
Safari
Edge

Is my data secure?

Yes, security measures include:

All data is encrypted in transit (HTTPS)
RxDB supports encryption at rest
API keys are stored securely in FlowState configuration, not in conversation data
Organization isolation ensures data separation between teams

Do I need to be online to use the Chat App?

For sending new messages, yes - you need connectivity to reach the LLM providers. However:

Your conversation history is stored locally in RxDB
You can view past conversations offline
When using LM Studio with local models, you can chat without internet
Data syncs when you reconnect

What are the message/token limits?

Limits depend on your chosen provider and model:

Provider	Model	Context Window
Anthropic	Claude Sonnet 4.5	200K tokens
Anthropic	Claude Opus 4.1	200K tokens
OpenAI	GPT-4o	128K tokens
OpenAI	GPT-4 Turbo	128K tokens
LM Studio	Varies	Depends on loaded model

How do I use LM Studio?

To use local models via LM Studio:

Download and install LM Studio
Download a model (e.g., Llama, Mistral, Phi)
Load the model in LM Studio
Start the local server (default: http://localhost:1234)
Select "LM Studio" as your provider in the Chat App

Billing and Usage

Do conversations count toward API usage?

Yes, messages sent to AI providers count toward your API usage with each provider:

Anthropic charges per token for Claude models
OpenAI charges per token for GPT models
LM Studio uses local resources (no API charges)

How can I reduce API costs?

To optimize costs:

Use faster/cheaper models when possible (Haiku, GPT-3.5 Turbo)
Keep prompts concise
Use LM Studio for development and testing
Review token usage displayed in message metadata

Still Have Questions?

If your question isn't answered here:

Check Troubleshooting for common issues
Review the Features documentation for detailed capabilities
Contact support for additional help