dexto
by truffle-ai·★ 620·Score 51
A TypeScript-based agent harness that includes MCP server capabilities for building and orchestrating AI applications.
Overview
Dexto is a comprehensive agent harness that can function as an MCP server through its stdio or HTTP/SSE transport modes. It provides a production-ready coding agent and supports building custom agents via YAML configuration. The system integrates with 50+ LLM models and offers MCP integration with 30+ tools. Dexto includes persistent sessions, memory management, and human-in-the-loop controls, making it suitable for building robust AI applications.
Try asking AI
After installing, here are 7 things you can ask your AI assistant:
When to choose this
Choose Dexto when you need a comprehensive agent framework for building autonomous coding agents and multi-agent systems with rich tool integration and persistent memory.
When NOT to choose this
Consider alternatives if you need a more focused MCP server for specific tools rather than a full agent framework, or if you prefer a different license as Dexto uses Elastic 2.0 which may have restrictions for some organizations.
Tools this server exposes
5 tools extracted from the READMEmcp--filesystem--read_fileRead file contents from the filesystem
mcp--filesystem--list_directoryList contents of a directory
agent-spawnerSpawn specialized sub-agents to handle complex subtasks
explore-agentFast, read-only codebase exploration agent
switchLLMSwitch between different language models during a conversation
Note: Tool names were extracted from documentation about MCP integration, permissions configuration, and agent spawning functionality. Some signatures were not explicitly documented.
Comparable tools
Installation
Installation
# macOS / Linux / WSL (native installer, recommended)
curl -fsSL https://dexto.ai/install | bash
# Windows PowerShell
irm https://dexto.ai/install.ps1 | iex
# Or build from source
git clone https://github.com/truffle-ai/dexto.git
cd dexto && pnpm install && pnpm install-cliRunning as MCP Server
# Start as MCP server
npx dexto --mode mcp --agent coding-agent --auto-approve --no-elicitationClaude Desktop Integration
Add to claude_desktop_config.json:
{
"mcpServers": {
"dexto": {
"command": "npx",
"args": ["-y", "dexto", "--mode", "mcp", "--agent", "coding-agent", "--auto-approve", "--no-elicitation"]
}
}
}FAQ
- How is Dexto different from other agent frameworks?
- Dexto provides both a coding agent out-of-the-box and the ability to build custom agents through YAML configuration. It has built-in MCP server capabilities and supports 50+ LLM models with instant switching.
- Can I use Dexto locally without sending data to cloud services?
- Yes, Dexto supports running with local models through Ollama or GGUF formats, keeping data on your machine with automatic GPU detection.
- How do I integrate Dexto with my existing workflow?
- Dexto provides multiple interfaces: CLI, Web UI, REST API, Discord, Telegram, and can be embedded in applications. You can also expose it as an MCP server to integrate with tools like Claude Code or Cursor.
On Hacker News
Recent discussion from the developer community.
- Story by shaunaks · 2025-10-28
Compare dexto with
Last updated · Auto-generated from public README + GitHub signals.