Skip to content
Evonic

Overview

Getting Started with Evonic

Section titled “Getting Started with Evonic”

Welcome to Evonic — a powerful, local-first platform for building and deploying AI agents. Whether you’re running a personal assistant, a multi-agent workflow, or a production deployment, Evonic gives you the tools to design, orchestrate, and manage agents at scale.

Evonic mascot

What is Evonic?

Section titled “What is Evonic?”

Evonic is an agentic AI platform built for open models. It lets you create AI agents powered by any OpenAI-compatible LLM — local, cloud, or anything in between. Agents can use tools, access knowledge bases, communicate across channels, and collaborate with each other.

Key Features

Section titled “Key Features”
  • Agent platform — design and deploy AI agents with custom models, tools, and knowledge bases
  • Multi-channel — connect agents to Telegram, WhatsApp (multi-agent, shared channels, concurrent numbers), Discord, and the web
  • Tool system — extend agent capabilities with tools, skills, plugins, and MCP servers
  • CMP & ATG — Context Memory Protocol and Atomic Task Graph for structured reasoning and reliable task execution
  • Evaluation — test and benchmark agent performance with expanded cross-session metrics and automated regression
  • Isolated sandbox — safe execution in Docker or high-performance Bubblewrap (bwrap) environments
  • Evonet — mesh network for multi-device agent deployment
  • OpenAI Codex integration — native support for Codex-optimized workflows with Chain-of-Thought thinking bubbles
  • Modern Panel UI — redesigned dashboard with modular panels, background process manager, and drag-and-drop file uploads

Quick Start

Section titled “Quick Start”
  1. Install Evonic — one-command install or manual setup
  2. Configure — set up your environment and model endpoints
  3. Create a Super Agent — set up the platform administrator
  4. Build your first agent — test it right away

What’s New in v1.0.0

Section titled “What’s New in v1.0.0”

Evonic v1.0.0 is a stable release that brings major architectural upgrades and a polished user experience. Here are the highlights:

Core Architecture

Section titled “Core Architecture”
  • CMP (Context Memory Protocol) — standardized agent memory management across all LLM providers, with three-tier memory: working, episodic, and semantic.
  • ATG (Atomic Task Graph) — complex goals are now decomposed into a DAG of atomic tasks with dependency tracking and parallel execution.
  • Bubblewrap Sandbox — the new bwrap workplace type offers near-instant, hardware-isolated sandboxes with zero daemon overhead.
  • Plugin Lifecycle System — formalized init → activate → active → deactivate lifecycle for reliable plugin management.

Agent Capabilities

Section titled “Agent Capabilities”
  • OpenAI Codex Integration — native support with Chain-of-Thought “thinking bubbles” that surface the agent’s internal reasoning.
  • WhatsApp Multi-Agent — shared channels, concurrent numbers, and multi-agent coordination within a single chat thread.
  • MCP Client Integration — connect to any Model Context Protocol server for instant access to a vast ecosystem of standardized tools.
  • Expanded Evaluator — cross-session metrics, automated regression testing, and dynamic prompt verification.

UI & Experience

Section titled “UI & Experience”
  • Panel UI — modular, resizable panels for multi-context workflows.
  • Background Process Manager — monitor and manage /detach jobs in real time.
  • Agent Detail Redesign — tabbed interface with live state viewer.
  • Drag-and-Drop Uploads — files dropped into chat are instantly available in the agent’s workplace.

New Tools & Commands

Section titled “New Tools & Commands”
  • send_channel_message, list_sessions, and Transcribe Audio tools
  • /dump — export session state as JSONL
  • /clear — now auto-archives sessions to prevent data loss
  • /unfocus — break out of stuck focus loops instantly

See the full Changelog for all details.

Next Steps

Section titled “Next Steps”