Dexto

Dexto is an AI agent platform backed by Y Combinator that lets developers deploy autonomous agents for coding, research, and task execution—locally or in the cloud. Agents maintain persistent memory across sessions, support human-in-the-loop approvals, and can work asynchronously on GitHub issues overnight. The platform supports 50+ LLMs including local models via Ollama, and exposes a CLI, Web UI, and API for flexible integration.

• Persistent Memory — Agents remember your codebase, patterns, and prior sessions via a built-in context store with vector-based RAG.
• Human-in-the-Loop Approvals — Review diffs and approve tool calls before anything ships, keeping you in full control of what gets deployed.
• 50+ LLM Support — Switch between cloud and local models (Ollama, llama.cpp) without changing your workflow; bring your own API keys (BYOK) or use Nova credits.
• Multi-Agent Collaboration — Spin up specialist agents for research, design, coding, and growth, then let them collaborate on complex tasks.
• Async GitHub Integration — Tag your cloud agent on a GitHub issue; it plans, codes, tests, and opens a PR ready for your morning review.
• MCP Server Support — Connect to multiple MCP servers to extend agent capabilities with external tools and data sources.
• Flexible Deployment — Run agents on your laptop, server, or Dexto Cloud; embed via CLI, Web UI, API, Slack, Discord, or your own app.
• Nova Gateway — Pay-as-you-go credit system for instant access to 50+ LLMs through a single endpoint without juggling API keys.

To get started, install the CLI with curl -fsSL https://dexto.ai/install | bash, connect your API keys or purchase Nova credits, and assign your first task or GitHub issue to an agent.