HYDRACODE - Multi-headed AI coding assistant with CLI, Telegram bot, MCP support, RAG for code and documents, and persistent memory
npm install hydracode-cliMulti-headed AI coding assistant with CLI, Telegram bot, MCP support, and persistent memory.
- Multi-Provider Support: OpenAI, Anthropic, Groq, Together, Ollama, LM Studio, and more
- Smart Router Mode: Automatically routes tasks to appropriate models based on complexity
- Telegram Bot: Use your AI assistant from anywhere via Telegram
- MCP Integration: Connect to Model Context Protocol servers for extended capabilities
- Persistent Memory: Remember conversations, preferences, and context across sessions
- Tool System: File operations, bash commands, search, and more
``bash`
npm install -g hydracode-cli
`bashRun the setup wizard
hydracode
Configuration
$3
On first run, HYDRACODE will guide you through setup:
1. Choose your LLM provider (OpenAI, Groq, Anthropic, etc.)
2. Enter your API key
3. Select a model$3
`bash
OpenAI
OPENAI_API_KEY=sk-...Groq (fast & free)
GROQ_API_KEY=gsk_...Anthropic
ANTHROPIC_API_KEY=sk-ant-...Or set via CLI
hydracode config --api-key YOUR_KEY --provider groq
`Commands
$3
- /help - Show available commands
- /clear - Clear conversation history
- /exit - Exit HYDRACODE$3
- /model - Change model & provider
- /bio - Set custom instructions for the AI
- /config - Show current configuration$3
- /memory - Show memory status
- /memory on - Enable persistent memory
- /memory off - Disable memory$3
- /mcp presets - List available server presets
- /mcp add brave-search - Add a preset server
- /mcp add myserver python server.py - Add custom server
- /mcp env server-name API_KEY=xxx - Set API keys
- /mcp connect - Connect to all servers
- /mcp tools - List available tools$3
- /gateway setup - Configure Telegram bot
- /serve - Start Telegram bot from CLI$3
- /routermode - Configure smart model routing
- Routes tasks to different models based on complexity:
- LOW: Simple questions → fast/cheap model
- MID: Moderate tasks → balanced model
- HIGH: Complex tasks → powerful modelMCP Server Presets
| Preset | Description | API Key |
|--------|-------------|---------|
|
brave-search | Web search | BRAVE_API_KEY |
| github | GitHub API | GITHUB_TOKEN |
| filesystem | File access | None |
| puppeteer | Browser automation | None |
| fetch | HTTP requests | None |
| sqlite | SQLite database | None |
| postgres | PostgreSQL | POSTGRES_CONNECTION_STRING |
| slack | Slack workspace | SLACK_BOT_TOKEN |Examples
$3
`bash
hydracode
> Create a Python web scraper that extracts headlines from news sites
`$3
`bash
GROQ_API_KEY=gsk_xxx hydracode
> Explain how React hooks work
`$3
`bash
hydracode
/gateway setup
(follow prompts to set up Telegram bot)
/serve
`$3
`bash
hydracode
/mcp add brave-search
/mcp env brave-search BRAVE_API_KEY=your_key
/mcp connect
> Search for the latest news about AI
``| Provider | Models | Notes |
|----------|--------|-------|
| OpenAI | gpt-4o, gpt-4o-mini, o1-preview | Best overall quality |
| Groq | llama-3.3-70b, mixtral | Fast & free tier |
| Anthropic | claude-3.5-sonnet, claude-3-opus | Great for coding |
| Together | Llama, Mistral, etc. | Many open models |
| Ollama | Any local model | Run models locally |
| LM Studio | Any local model | Local with GUI |
MIT