agnt-gg/agnt

The local-first operating system for building, running, and improving AI agents, workflows, and autonomous goals.

What it solves

AGNT provides a durable, local-first runtime for AI agents that moves beyond disposable chat interfaces. It solves the problem of AI agents lacking persistent state, repeatable workflows, and the ability to evolve through evaluation and memory across different sessions.

How it works

AGNT operates as a desktop application or self-hosted backend (using Electron, Vue.js, and SQLite) that integrates three primary execution modes:

  • Agents: Persistent workers for reasoning, tool use, and collaboration.
  • Workflows: A visual DAG (Directed Acyclic Graph) engine for repeatable automation with over 60 nodes.
  • Goals: An autonomous system for planning, executing, and evaluating long-term objectives.

These modes are interconnected; for example, a goal can trigger a workflow, which in turn can call a specialized agent. The system also includes "Forge" tools to AI-generate agents, workflows, and tools, and a "SkillForge" to turn execution traces into reusable skills.

Who it’s for

It is designed for developers and power users who want local ownership of their AI data, need to orchestrate complex multi-agent systems, or require AI automations that run independently of a chat window (e.g., via messaging triggers or scheduled tasks).

Highlights

  • Local-first architecture: Data is stored locally in SQLite and the filesystem.
  • Multi-provider support: Compatible with 15+ AI providers including OpenAI, Anthropic, Gemini, and local endpoints.
  • Extensibility: Supports MCP (Model Context Protocol) servers, custom plugins (.agnt packages), and a local API on port 3333.
  • Subagent Delegation: Native ability to delegate work across agents, workflows, and async tools in parallel.
  • Self-Improving: Uses execution traces and evaluations to extract insights and create reusable skills.

Related

  • Dispatch
    ChatGPT Work: A Technical Deep Dive into OpenAI's Agentic Workflow Tool

    ChatGPT Work is a paid agentic platform that extends standard chat with a persistent filesystem, headless Chrome browser, and internet-enabled code execution to automate complex, multi-step tasks.

  • Project
    openairymax/agentrt

    AgentRT is an OS-level runtime infrastructure for AI agents, providing a microkernel-style foundation for cognitive loops, memory layering, and secure inter-process communication.

  • Project
    hAcKlyc/MyAgents

    An open-source desktop workspace for personal AI agents that integrates local files, terminals, and a task scheduling system to move beyond simple chat interfaces.

  • Project
    wecode-ai/Wegent

    An open-source AI work system that integrates a local desktop workbench with cloud agents and remote machines for collaborative AI-driven coding and task execution.

  • Project
    Lamatic/AgentKit

    AgentKit is an open‑source SDK that provides a catalog of 70 ready‑made “kits” (full apps, pipelines, or templates) for building, testing and deploying AI agents. Kits are defined by a simple config file, can be edited in a visual flow studio, and are shipped as Next.js apps that run server‑less. The collection covers support triage, code review, RAG chat, hiring assistants, legal/medical bots, and many other business‑focused agents, making it a turnkey platform for creating reliable, enterprise‑grade AI agents.