Free
$0
- 30 trial messages, then 50 / month
- 1 AI provider
- Bring your own API key
- On-device chats & Keychain
- Streaming Markdown, LaTeX & code
- Themes & basic parameters
- Built-in prompt templates
Privacy-first AI chat
Bring your own key (BYOK)
Connect any model, local or cloud
- Multi-Provider Support: Connect to any model via OpenAI/Anthropic-compatible APIs, including local models
- Chat Presets: Save prompt, model, parameter, and tool as reusable presets. Switch between them with one tap to boost your workflow efficiency
- MCP protocol support: MCP integration for third-party tool calling
- Prompt Variables: Dynamic placeholders for context-aware prompts
- Live Streaming: Real-time typewriter animation rendering with LaTeX formulas and Markdowns
- Local-First Storage: Keys in Apple Keychain, chats in on-device SwiftData — no account, no middleware
Download on the
App Store
Coming soon
Mac App Store
Start a Conversation
Choose a template, or send a message to start from scratch.
Chat Preset
Daily Talk
Research
Translator
Tap a preset to load model, prompt & tools
How to choose the best mortgage plan in the US in 2026 — fixed vs. variable?
Reasoning
0.5s
The right plan depends on how long I'll stay in the home. I'll pull today's rates from Freddie Mac, then run the amortization on a $500k loan to compare monthly payment and total interest for a 30-year fixed, 15-year fixed, and 5/1 ARM.
web_search
1.2s
{ "query": "current US mortgage rates August 2026", "max_results": 5 }Freddie Mac PMMS · 30-yr 6.69% · 15-yr 6.01% · ARM 6.51%
| Plan | Rate | Payment | Interest |
|---|---|---|---|
| 30-yr fixed | 6.69% | $3,223 | $660k |
| 15-yr fixed | 6.01% | $4,222 | $260k |
| 5/1 ARM | 6.51% | $3,163 | $639k |
Monthly payment on a $500k loan:
M = P·r(1+r)ⁿ / ((1+r)ⁿ − 1)
Amortization in Python:
python
Copy
Type a message...
Chat Presets · Live streaming · LaTeX · Tables · Code
Bring Your Own Key
Any model. One place.
- OpenAI, Anthropic, Gemini, DeepSeek, Mistral — or any OpenAI-compatible endpoint
- Bring your own key: stored in Apple Keychain, never on our servers
- Auto-fetch available models from any base URL
- Every chat stays on your device with SwiftData
Add Provider
BYOK
Choose a provider
OpenAI
api.openai.com/v1
Anthropic
api.anthropic.com
Google Gemini
generativelanguage.googleapis.com
DeepSeek
api.deepseek.com
OpenAI Compatible
Any base URL
+ OpenRouter, Mistral, Groq, LM Studio, and more
OpenAI Compatible
Base URL
http://localhost:11434
Ollama
API Key
Optional · not required
Fetched 3 models
Ollama · localhost:11434
Models
qwen3.6:35b-a3b35B · 3B active
llama3.3:70bGeneral purpose
deepseek-r1:32bReasoning
Chat Preset
One tap, a different expert.
- Save model, prompt, parameters, and tools as a named preset
- Flip from Research to Translator in one tap — no reconfiguring
- Prompt variables — {{current_date}}, {{current_time}}, {{model_id}} resolved when you send
Chat Preset
1 Tap
Presets
Prompt
"Act as a research analyst. Cite sources and flag uncertainty."
Variables
{{current_date}} → 2026-08-13
Model
gpt-4o
Temperature
0.4
Tools
web_search
fetch
Model · prompt · params · tools
+ New Preset
MCP Tools
Plug your AI into the real world.
- Connect any MCP server — hundreds of ready-made tools, or your own
- Bearer-token and custom-header authentication
- Validate the connection and auto-discover available tools
MCP
JSON-RPC
Server Details
Name
web-tools
URL
https://tools.local/mcp
Example
Transport
Streamable HTTP · SSE
Authentication
Type
Bearer Token
Bearer Token
••••••••••••••••
Connect
Connected to web-tools
Fetched 3 tools · protocol 2026-07-28
Tools
web_search_exaExa · Neural web search
web_fetch_exaExa · Fetch as clean markdown
batch_searchAnySearch · Parallel queries
Who won the 2026 Fields Medal?
Parameters
Fine-tune it all, in your own way.
- Temperature, Top P, Top K, Min P, Max Tokens
- Reasoning effort: Low / Medium / High for supported models
- Presence & frequency penalty, seed, stop sequences
- Or drop in provider-specific JSON when you need to go deeper
Session Parameters
Fine control
Temperature
0.7
Top P
0.9
Max Tokens
2048
Reasoning Effort
LowMediumHigh
Presence Penalty
0.2
Frequency Penalty
0.5
Stop Sequences
</output>
Per session or preset
Provider JSON
Appearance
Make it yours, native SwiftUI.
- Light, dark, or auto — match your system
- Nine accent colors, or pick any custom color
- Adjustable font size: small, medium, or large
- Built with SwiftUI — feels at home on iPhone and iPad (Mac coming soon)
Appearance
Settings
Mode
Accent
Font size
Pricing
Free to try. Pro for life.
Lifetime
Pro
$19.99
- Everything in Free, plus:
- Unlimited messages & providers
- Chat Presets
- MCP tools
- Custom prompt templates
- Advanced parameters
- Bulk export history
FAQ
Common questions
Do you store my chats or API keys?
No. API keys stay in Apple Keychain. Conversations stay on your device. BayesChat has no account system and no cloud that receives your chats.
What does BYOK mean?
Bring your own key. You connect providers with your own API key and pay them directly. We do not resell tokens or sit in the middle of your requests.
Which devices and providers are supported?
Available now on iPhone and iPad; Mac is coming soon. Use OpenAI, Anthropic, Gemini, OpenRouter, DeepSeek, Mistral, LM Studio, Ollama, or any OpenAI-compatible endpoint with a custom URL.
Is Pro a subscription?
No. BayesChat is free to download. Pro is a one-time in-app purchase with lifetime access — no subscription.
What is included in Free vs Pro?
Free includes BYOK, on-device storage, streaming Markdown/LaTeX, themes, and basic parameters, with message and provider limits. Pro unlocks unlimited messages and providers, Chat Presets, MCP tools, custom prompts, advanced parameters, and bulk export.
What is MCP?
Model Context Protocol. You can add MCP servers so the model can use external tools such as web search, fetch, or your own JSON-RPC tools — enabled per chat.
Contact Us
Questions or feedback?
Email us about bugs, feature requests, or anything else. We read every message.