I’ve been building something I’ve personally needed for a long time: a mobile AI companion that isn’t tied to one model or one provider. Most assistants lock you into a single LLM and a single way of reasoning. I wanted something more flexible, more composable, and more portable.
So I built Bubs, a cross-platform (Android + iOS) AI companion with multi-LLM orchestration and a local-first automation engine.
Core idea
Bubs can talk to five LLMs (more coming.):
Anthropic Claude
OpenAI GPT-4
Google Gemini
Cohere Command
Local/on-device models
It uses a routing engine that selects the best model for a task based on quality, speed, or cost. It also supports ensemble mode, which queries multiple providers and aggregates the results for higher-confidence reasoning. If a provider is down, it falls back automatically.
Chat + automation
Beyond a chat UI, Bubs includes a natural language workflow automation system (similar in spirit to Zapier, but embedded locally):
Create flows with natural language
Edit visually
Run triggers & actions
Connectors for Slack, GitHub, Google, Notion, Email, Webhooks
Execution history
All on-device, no backend required
The automations are intended to let users build personal “micro-agents” without needing cloud infrastructure.
Local-first architecture
Bubs stores API keys securely (Keychain / EncryptedStore) and doesn’t require a backend to function. Your prompts and automations stay on your device unless you explicitly send them to a provider. This also eliminates ongoing hosting costs.
I’m finishing debugging the Android build now and polishing the iOS build next. Once both are stable, I’m planning a small early-access presale for people who want to test it while I continue refining the automation engine and routing logic.
Why I’m posting
I’d love feedback from the HN community on:
Multi-model orchestration strategies
Ideas for additional connectors or triggers
Local privacy concerns or improvements
Whether others are working on similar architectures
UI/UX considerations for flow creation on mobile
Suggestions for open-source components Bubs could integrate with
This is a solo build, and I’m trying to make the architecture as clean and extensible as possible. Happy to answer technical questions.
Thanks for taking a look.
Bubs is always with you.
I’ve been building something I’ve personally needed for a long time: a mobile AI companion that isn’t tied to one model or one provider. Most assistants lock you into a single LLM and a single way of reasoning. I wanted something more flexible, more composable, and more portable.
So I built Bubs, a cross-platform (Android + iOS) AI companion with multi-LLM orchestration and a local-first automation engine.
Core idea
Bubs can talk to five LLMs (more coming.):
Anthropic Claude
OpenAI GPT-4
Google Gemini
Cohere Command
Local/on-device models
It uses a routing engine that selects the best model for a task based on quality, speed, or cost. It also supports ensemble mode, which queries multiple providers and aggregates the results for higher-confidence reasoning. If a provider is down, it falls back automatically.
Chat + automation
Beyond a chat UI, Bubs includes a natural language workflow automation system (similar in spirit to Zapier, but embedded locally):
Create flows with natural language
Edit visually
Run triggers & actions
Connectors for Slack, GitHub, Google, Notion, Email, Webhooks
Execution history
All on-device, no backend required
The automations are intended to let users build personal “micro-agents” without needing cloud infrastructure.
Local-first architecture
Bubs stores API keys securely (Keychain / EncryptedStore) and doesn’t require a backend to function. Your prompts and automations stay on your device unless you explicitly send them to a provider. This also eliminates ongoing hosting costs.
Tech stack
Android: Jetpack Compose, MVVM, Room, DataStore, Hilt iOS: SwiftUI, MVVM, Keychain, CloudKit-ready architecture
State of the project
I’m finishing debugging the Android build now and polishing the iOS build next. Once both are stable, I’m planning a small early-access presale for people who want to test it while I continue refining the automation engine and routing logic.
Why I’m posting
I’d love feedback from the HN community on:
Multi-model orchestration strategies
Ideas for additional connectors or triggers
Local privacy concerns or improvements
Whether others are working on similar architectures
UI/UX considerations for flow creation on mobile
Suggestions for open-source components Bubs could integrate with
This is a solo build, and I’m trying to make the architecture as clean and extensible as possible. Happy to answer technical questions.
Thanks for taking a look. Bubs is always with you.
– J