dexto

<a href="https://dexto.ai"> <img src="https://raw.githubusercontent.com/truffle-ai/dexto/main/.github/assets/dexto-logo.svg" alt="Dexto" width="100%" style="max-width: 1000px" /> </a> <p align="center"> <img src="https://img.shields.io/badge/Status-Beta-yellow"> <img src="https://img.shields.io/badge/License-Elastic%202.0-blue.svg"> <a href="https://discord.gg/GFzWFAAZcm"><img src="https://img.shields.io/badge/Discord-Join%20Chat-7289da?logo=discord&logoColor=white"></a> <a href="https://deepwiki.com/truffle-ai/dexto"><img src="https://deepwiki.com/badge.svg"></a> </p> **An all-in-one toolkit to build agentic applications that turn natural language into real-world actions.** <div align="center"> <img src="https://github.com/user-attachments/assets/97a09f5b-6f7d-4d4b-9b3f-3c87fd964cbe" alt="Dexto Demo" width="700" /> </div> ## What is Dexto? Dexto is a universal agent interface for building agentic apps—software that understands natural language and takes real-world actions. It orchestrates LLMs, tools, and data into persistent, stateful systems with memory, so you can rapidly create AI assistants, copilots, and context-aware apps that think, act and feel alive. #### With Dexto, you can build: - **Autonomous Agents** - Agents that plan, execute, and adapt to user goals. - **Digital Companions** - AI assistants & copilots that remember context and anticipate needs. - **Multi-Agent Systems** - Architect agents that collaborate, delegate, and solve complex tasks together. - **MCP Clients** - Connect multiple tools, files, APIs, and data via MCP Servers. - **Agent-as-a-Service** – Transform your existing SaaS products and APIs into dynamic, conversational experiences. - **Agentic Applications** – Integrate Dexto as a reasoning engine to power interactive, multimodal, AI-native applications. <!-- Dexto unifies a configuration-driven framework, robust runtime, and seamless developer experience—so you can build, deploy, and iterate with ease. - **Framework** – Define agent behavior in YAML. Instantly swap models and tools without touching code. - **Runtime** – Execution with session management, conversation memory, and multimodal support. - **Interfaces & Tooling** – Native support for CLI, Web, APIs, and a TypeScript SDK. Develop, debug, and interact with your agent across any platforms from day one. --> ### What You Get - **Batteries Included** – Session management, tool orchestration, multimodal support. - **20+ LLMs** – Instantly switch between OpenAI, Anthropic, Google, Groq, local models or bring your own. - **Run Anywhere** – Local for privacy, cloud for reach, or hybrid. *Same agent, any deployment.* - **Native Multimodal** – Text, images, files, and tools in a single conversation. *Upload screenshots, ask questions, take actions.* - **Persistent Sessions** – Conversations, context, and memory are saved and can be exported, imported, or shared across environments. - **Flexible Interfaces** – One agent, endless ways to interact: Ready to-use CLI, WebUI, APIs, or integrate with your own UI. - **Production Ready** – Observability and error handling built-in. - **Tooling & MCP** – Integrate 100+ tools and connect to external servers via the Model Context Protocol (MCP). - **Customizable Agents** – Define agent behavior, tools, and prompts in YAML or TypeScript. - **Pluggable Storage** – Use Redis, PostgreSQL, SQLite, in-memory, and more for cache and database backends. --- ## Installation ```bash # NPM global npm install -g dexto # —or— build from source # this sets up dexto CLI from the cloned code git clone https://github.com/truffle-ai/dexto.git cd dexto && pnpm install && pnpm install-cli ``` ## Quick Start ```bash # 1. Run setup workflow - this prompts for your preferred LLM and API keys and starts the interactive CLI dexto # 2. Try a multi-step task dexto "create a snake game in HTML/CSS/JS, then open it in the browser" # 3. Launch the Dexto Web UI dexto --mode web ``` In 2 -> Dexto will use filesystem tools to write code and browser tools to open it — all from a single prompt. The Web UI allows you to navigate previous conversations and experiment with different models, tools and more. ### Skip Tool Prompts While Prototyping ```bash dexto --auto-approve "refactor my project using the filesystem and browser tools" ``` Use the `--auto-approve` flag to bypass confirmation prompts when you trust the tools being invoked—perfect for fast local iteration. Remove the flag when you want explicit approval again. ## Agent Recipes Dexto comes with pre-built agent recipes for common use cases. Install and use them instantly: ```bash # List available agents dexto list-agents # Install specific agents dexto install nano-banana-agent podcast-agent # Use an agent dexto --agent nano-banana-agent "create a futuristic cityscape with flying cars" dexto --agent podcast-agent "generate a podcast intro with two hosts discussing AI" ``` **Available Agents:** - **Nano Banana Agent** – Advanced image generation and editing using Google's Nano Banana (Gemini 2.5 Flash Image) - **Podcast Agent** – Advanced podcast generation using Google Gemini TTS for multi-speaker audio content - **Database Agent** – Demo agent for SQL queries and database operations - **Image Editor Agent** – Image editing and manipulation - **Music Agent** – Music creation and audio processing - **PDF Agent** – Document analysis and conversation - **Product Researcher** – Product naming and branding research - **Triage Agent** – Demo multi-agent customer support routing system Each agent is pre-configured with the right tools, prompts, and LLM settings for its domain. No setup required—just install and start building. More ready-to-run recipes live in [`agents/`](agents/) and the [docs site](https://docs.dexto.ai/). ## Examples & Demos ### 🛒 Browser Agent: Amazon Shopping Assistant **Task:** `Can you go to amazon and add some snacks to my cart? I like trail mix, cheetos and maybe surprise me with something else?` ```bash # Default agent has browser tools dexto ``` <a href="https://youtu.be/C-Z0aVbl4Ik"> <img src="https://github.com/user-attachments/assets/3f5be5e2-7a55-4093-a071-8c52f1a83ba3" alt="Dexto: Amazon shopping agent demo" width="600"/> </a> ### 👁️ Computer Vision: Face Detection & Annotation **Task:** `Detect all faces in this image and draw bounding boxes around them.` ```bash dexto --agent image-editor-agent ``` <img src="https://github.com/user-attachments/assets/7e4b2043-c39a-47c7-a403-a9665ee762ce" alt="Face Detection Demo" width="600"> ### 🎙️ Podcast Agent: Generate AI Podcast Intros **Task:** `Generate an intro for a podcast about the latest in AI.` ```bash dexto --agent podcast-agent ``` <img src="https://github.com/user-attachments/assets/cfd59751-3daa-4ccd-97b2-1b2862c96af1" alt="Podcast Agent Demo" width="600"/> ### 🖼️ Hugging Face: Image Generation **Task:** `Generate a photo of a baby panda.` <img src="https://github.com/user-attachments/assets/570cbd3a-6990-43c5-b355-2b549a4ee6b3" alt="Hugging Face Image Generation Demo" width="600"/> ### 🛠️ Adding Custom MCP Servers You can add your own Model Context Protocol (MCP) servers to extend Dexto's capabilities with new tools or data sources. Just edit your agent YAML or add it directly in the WebUI. <img src="https://github.com/user-attachments/assets/1a3ca1fd-31a0-4e1d-ba93-23e1772b1e79" alt="Add MCP Server Example" width="600"/> ### 🎮 Web App Development: Snake Game **Task:** `Create a snake game in HTML/CSS/JS, then open it in the browser` ```bash dexto "create a snake game in HTML/CSS/JS, then open it in the browser" ``` <img src="https://github.com/user-attachments/assets/6901ba82-3c2a-4177-baf0-7d44d04a5f27" alt="Snake Game Development Demo" width="600"/> ### 📧 Send Email Summaries to Slack **Task:** `Summarize emails and send highlights to Slack` ```bash dexto --agent ./agents/examples/email_slack.yml ``` <img src="assets/email_slack_demo.gif" alt="Email to Slack Demo" width="600"> --- ## Run Modes | Mode | Command | Best for | |------|---------|----------| | **Interactive CLI** | `dexto` | Everyday automation & quick tasks | | **Web UI** | `dexto --mode web` | Friendly chat interface w/ image support | | **Headless Server** | `dexto --mode server` | REST & WebSocket APIs for agent interaction | | **MCP Server (Agent)** | `dexto --mode mcp` | Exposing your agent as a tool for others via stdio | | **MCP Server (Aggregator)** | `dexto mcp --group-servers` | Re-exposing tools from multiple MCP servers via stdio | | **Discord Bot** | `dexto --mode discord` | Community servers & channels ([Requires Setup](packages/cli/src/discord/README.md)) | | **Telegram Bot** | `dexto --mode telegram` | Mobile chat ([Requires Setup](packages/cli/src/telegram/README.md)) | Run `dexto --help` for **all flags, sub-commands, and environment variables**. ## Configuration ### Agent Configuration Dexto treats each configuration as a unique agent allowing you to define and save combinations of LLMs, servers, storage options, etc. based on your needs for easy portability. Define agents in version-controlled YAML. Change the file, reload, and chat—state, memory, and tools update automatically. ```yaml # agents/my-agent.yml llm: provider: openai model: gpt-4.1-mini apiKey: $OPENAI_API_KEY mcpServers: filesystem: type: stdio command: npx args: ['-y', '@modelcontextprotocol/server-filesystem', '.'] web: type: stdio command: npx args: ['-y', '@modelcontextprotocol/server-brave-search'] systemPrompt: | You are a helpful AI assistant with access to files and web search. ``` ### LLM Providers Switch between providers instantly—no code changes required. | Provider | Models | Setup | |----------|--------|-------| | **OpenAI** | `gpt-5`, `gpt-5-mini`, `gpt-5-nano`, `gpt-4.1`, `gpt-4.1-mini`, `gpt-4.1-nano`, `gpt-4o`, `gpt-4o-mini`, `gpt-4o-audio-preview`, `o4-mini`, `o3`, `o3-mini`, `o1` | `export OPENAI_API_KEY=...` | | **Anthropic** | `claude-sonnet-4-5-20250929`, `claude-opus-4-1-20250805`, `claude-4-opus-20250514`, `claude-4-sonnet-20250514`, `claude-3-7-sonnet-20250219`, `claude-3-5-sonnet-20240620`, `claude-3-5-haiku-20241022` | `export ANTHROPIC_API_KEY=...` | | **Google** | `gemini-2.5-pro`, `gemini-2.5-flash`, `gemini-2.5-flash-lite`, `gemini-2.0-flash`, `gemini-2.0-flash-lite` | `export GOOGLE_GENERATIVE_AI_API_KEY=...` | | **Groq** | `llama-3.3-70b-versatile`, `meta-llama/llama-4-scout-17b-16e-instruct`, `meta-llama/llama-4-maverick-17b-128e-instruct`, `qwen/qwen3-32b`, `gemma-2-9b-it`, `openai/gpt-oss-20b`, `openai/gpt-oss-120b`, `moonshotai/kimi-k2-instruct`, `deepseek-r1-distill-llama-70b` | `export GROQ_API_KEY=...` | | **xAI** | `grok-4`, `grok-3`, `grok-3-mini`, `grok-code-fast-1` | `export XAI_API_KEY=...` | | **Cohere** | `command-a-03-2025`, `command-r-plus`, `command-r`, `command-r7b` | `export COHERE_API_KEY=...` | ```bash # Switch models via CLI dexto -m claude-sonnet-4-5-20250929 dexto -m gemini-2.5-pro ``` See our [Configuration Guide](https://docs.dexto.ai/docs/category/guides/) for complete setup instructions. ## Programmatic API Install the `@dexto/core` library, and build applications with the `DextoAgent` class. Everything the CLI can do, your code can too. ```bash npm install @dexto/core ``` ```ts import { DextoAgent } from '@dexto/core'; // Create and start agent const agent = new DextoAgent({ llm: { provider: 'openai', model: 'gpt-4.1-mini', apiKey: process.env.OPENAI_API_KEY } }); await agent.start(); // Run tasks const response = await agent.run('List the 5 largest files in this repo'); console.log(response); // Hold conversations await agent.run('Write a haiku about TypeScript'); await agent.run('Make it funnier'); await agent.stop(); ``` See our [TypeScript SDK docs](https://docs.dexto.ai/api/category/typescript-sdk/) for complete examples with MCP tools, sessions, and advanced features. --- ## Advanced Usage ### Session Management Create and manage multiple conversation sessions with persistent storage. ```typescript const agent = new DextoAgent(config); await agent.start(); // Create and manage sessions const session = await agent.createSession('user-123'); await agent.run('Hello, how can you help me?', undefined, 'user-123'); // List and manage sessions const sessions = await agent.listSessions(); const sessionHistory = await agent.getSessionHistory('user-123'); await agent.deleteSession('user-123'); // Search across conversations const results = await agent.searchMessages('bug fix', { limit: 10 }); ``` ### LLM Management Switch between models and providers dynamically. ```typescript // Get current configuration const currentLLM = agent.getCurrentLLMConfig(); // Switch models (provider inferred automatically) await agent.switchLLM({ model: 'gpt-4.1-mini' }); await agent.switchLLM({ model: 'claude-sonnet-4-5-20250929' }); // Switch model for a specific session id 1234 await agent.switchLLM({ model: 'gpt-4.1-mini' }, '1234') // Get supported providers and models const providers = agent.getSupportedProviders(); const models = agent.getSupportedModels(); const openaiModels = agent.getSupportedModelsForProvider('openai'); ``` ### MCP Manager For advanced MCP server management, use the MCPManager directly. ```typescript import { MCPManager } from '@dexto/core'; const manager = new MCPManager(); // Connect to MCP servers await manager.connectServer('filesystem', { type: 'stdio', command: 'npx', args: ['-y', '@modelcontextprotocol/server-filesystem', '.'] }); // Access tools, prompts, and resources const tools = await manager.getAllTools(); const prompts = await manager.getAllPrompts(); const resources = await manager.getAllResources(); // Execute tools const result = await manager.executeTool('readFile', { path: './README.md' }); await manager.disconnectAll(); ``` ### Storage & Persistence Configure storage backends for production-ready persistence and caching. ```yaml # agents/production-agent.yml storage: cache: type: redis url: $REDIS_URL maxConnections: 100 database: type: postgres connectionString: $POSTGRES_CONNECTION_STRING maxConnections: 25 sessions: maxSessions: 1000 sessionTTL: 86400000 # 24 hours ``` **Supported Backends:** - **Cache**: Redis, In-Memory (fast, ephemeral) - **Database**: PostgreSQL, SQLite, In-Memory (persistent, reliable) **Use Cases:** - **Development**: In-memory for quick testing - **Production**: Redis + PostgreSQL for scale - **Simple**: SQLite for single-instance persistence See the [DextoAgent API Documentation](https://docs.dexto.ai/api/dexto-agent/) for complete method references. --- ## CLI Reference <details> <summary>Click to expand for full CLI reference (`dexto --help`)</summary> ``` Usage: dexto [options] [command] [prompt...] Dexto CLI allows you to talk to Dexto, build custom AI Agents, build complex AI applications like Cursor, and more. Run dexto interactive CLI with `dexto` or run a one-shot prompt with `dexto -p "<prompt>"` or `dexto "<prompt>"` Start with a new session using `dexto --new-session [sessionId]` Run dexto web UI with `dexto --mode web` Run dexto as a server (REST APIs + WebSockets) with `dexto --mode server` Run dexto as a discord bot with `dexto --mode discord` Run dexto as a telegram bot with `dexto --mode telegram` Run dexto agent as an MCP server with `dexto --mode mcp` Run dexto as an MCP server aggregator with `dexto mcp --group-servers` Check subcommands for more features. Check https://github.com/truffle-ai/dexto for documentation on how to customize dexto and other examples Arguments: prompt Natural-language prompt to run once. If not passed, dexto will start as an interactive CLI Options: -v, --version output the current version -a, --agent <name|path> Agent name or path to agent config file -p, --prompt <text> One-shot prompt text. Alternatively provide a single quoted string as positional argument. -s, --strict Require all server connections to succeed --no-verbose Disable verbose output --no-interactive Disable interactive prompts and API key setup -m, --model <model> Specify the LLM model to use -r, --router <router> Specify the LLM router to use (vercel or in-built) --new-session [sessionId] Start with a new session (optionally specify session ID) --mode <mode> The application in which dexto should talk to you - cli | web | server | discord | telegram | mcp (default: "cli") --web-port <port> optional port for the web UI (default: "3000") --no-auto-install Disable automatic installation of missing agents from registry -h, --help display help for command Commands: create-app Scaffold a new Dexto Typescript app init-app Initialize an existing Typescript app with Dexto setup [options] Configure global Dexto preferences install [options] [agents...] Install agents from the registry uninstall [options] [agents...] Uninstall agents from the local installation list-agents [options] List available and installed agents which <agent> Show the path to an agent mcp [options] Start Dexto as an MCP server. Use --group-servers to aggregate and re-expose tools from configured MCP servers. ``` </details> --- ## Next Steps * **[Quick Start](https://docs.dexto.ai/docs/getting-started/intro/)** – Get up and running in minutes. * **[Configuration Guide](https://docs.dexto.ai/docs/category/guides/)** – Configure agents, LLMs, and tools. * **[Building with Dexto](https://docs.dexto.ai/docs/category/tutorials/)** – Developer guides and patterns. * **[API Reference](https://docs.dexto.ai/api/)** – REST APIs, WebSocket, and SDKs. --- ## Telemetry We collect anonymous usage data (no personal/sensitive info) to help improve Dexto. This includes: - Commands used - Command execution time - Error occurrences - System information (OS, Node version) - LLM Models used To opt-out: Set env variable `DEXTO_ANALYTICS_DISABLED=1` ## Contributing We welcome contributions! Refer to our [Contributing Guide](./CONTRIBUTING.md) for more details. ## Community & Support Dexto is built by the team at [Truffle AI](https://www.trytruffle.ai). Join our Discord to share projects, ask questions, or just say hi! [](https://discord.gg/GFzWFAAZcm) If you enjoy Dexto, please give us a ⭐ on GitHub—it helps a lot! <div align="left"/> [](https://x.com/intent/user?screen_name=Road_Kill11) [](https://x.com/intent/user?screen_name=shaun5k_) </div> --- ## Contributors Thanks to all these amazing people for contributing to Dexto! [](https://github.com/truffle-ai/dexto/graphs/contributors) --- ## License Elastic License 2.0. See [LICENSE](LICENSE) for full terms.