Convergent is a lightweight multi-model AI chat application that lets users query multiple LMs simultaneously, compare outputs side-by-side, generate synthesized responses, search the web for real-time information, and organize conversations locally.
Unlike traditional AI chat apps that rely on a single response, Convergent helps users validate outputs across multiple models and current web sources.
Built as a single HTML application with no framework dependency.
Most AI chat apps work like this:
- Ask one model
- Get one response
- Hope it's correct
Convergent improves that workflow:
- Query multiple models simultaneously
- Compare responses side-by-side
- Generate synthesized answers
- Search the web for current information
- Save chats securely
- Organize conversations efficiently
Send one prompt to multiple AI models at the same time and compare responses in parallel.
Generate a final answer by combining outputs from multiple model responses.
Search the web for real-time information directly inside chats.
Useful for:
- Latest news
- Current events
- Recent documentation
- Real-time research
- Fact verification
Supports configurable search providers.
Use built-in slash commands for faster workflows.
Examples:
/summarize/rewrite/explain/debug
Users can also create custom slash commands for repetitive workflows. Type / in the prompt. See list of available slash commands or create new.
All conversations remain stored on your device.
No mandatory cloud storage.
Protect conversations using passphrase-based encryption.
Group chats into projects for better long-term organization.
Search across previous chats instantly.
Backup and restore conversations easily.
- Download the HTML file
- Open it in your browser
- Add your API keys
- Start chatting
No backend setup required.
No framework installation required.
Convergent works with any provider that exposes an API endpoint.
Examples:
- OpenAI
- Anthropic
- Google Gemini
- OpenRouter
- Groq
- Together AI
- Perplexity
- DeepSeek
- Mistral
- Ollama
- LM Studio
- Self-hosted OpenAI-compatible endpoints
- Custom API providers
If you're using Ollama, allow browser-based requests by setting OLLAMA_ORIGINS.
Windows:
set OLLAMA_ORIGINS=*Linux (bash) and Mac (Terminal):
export OLLAMA_ORIGINS=*To access nvidia models, install nodejs in your system and run node convergent-proxy.js command and leave the terminal running. If you close this terminal Nvidia models wont work. In the html, settings/Endpoints option, enter http://localhost:3030 url and test proxy connection. Enter Nvidia API URL, key and access the models.
NVIDIA’s APIs expect requests to come from controlled environments (server, authenticated clients) and often reject raw fetch calls from frontend HTML apps.
The local proxy script runs on your machine (e.g., Node.js server) and forwards requests with proper headers, API‑key handling, and sometimes rewritten origin, so NVIDIA’s backend sees them as allowed.
This is only needed when accessing Nvidia models.
Use your preferred search provider for real-time information retrieval.
Examples:
- Tavily
- SearXNG
- Brave Search
- SerpAPI
- Serper
- Exa
- Custom search APIs
- HTML
- CSS
- JavaScript
- LocalStorage
- IndexedDB
- Browser Crypto APIs
Single-file architecture.
If you want to rename your existing storage folder or move your chats/configurations to a new folder:
- Use the Export Data option on the home page directly or navigate to Settings → Data
- Use Export Data to create a backup
- Rename your existing folder or create a new folder
- Use the Import Data option on the home page directly or navigate to Settings → Data
- Select the new folder location
- Import your exported backup
- Enter your existing password/passphrase
Your chats, configurations, and encrypted data will be restored and accessible in the new folder.
This project is licensed under the MIT License - see the LICENSE file for details