evonic
Open Agentic AI Platform - The home your agents deserve
What is it?
What it is
Evonic is an agentic AI framework for designing, building, and orchestrating intelligent agents from concept to production, allowing users to define models, tools, knowledge bases, channels, and skills and compose them into multi-agent systems.
Why it exists
To provide a safe, extensible platform for autonomous agent operations across diverse environments, enabling multi-agent swarms, workplace-agnostic execution, and real-time heuristic safety checks to prevent harmful actions.
Who should use it
Developers, DevOps engineers, AI researchers, and product teams looking to build, orchestrate, and deploy multi‑agent systems with custom tools, knowledge bases, and secure execution across distributed environments.
Who should avoid it
Non‑technical users, small teams without access to an OpenAI‑compatible LLM endpoint, or those who prefer simple chatbot solutions without the need for agent swarms or advanced safety features.
How it works
A quick walkthrough in plain English
How evonic works
Step 1 of 3
You interact with it
Open evonic, send a request, or connect it to your stack.
Features
Advantages
- Supports true multi‑agent orchestration with native communication
- Provides safety via real‑time heuristic checks and injection guard
- Allows agents to run anywhere (local, SSH, tunnel) with a single abstraction
- Reduces LLM costs with RTK‑based token compression
- Offers modular skills and plugins for extensibility
- Built‑in scheduler for recurring tasks
- Integrated channels (Telegram, WhatsApp, Discord, Web UI)
- Automated evaluation engine for testing LLM outputs
- Easy deployment via one‑liner install and Docker sandbox
- Full backup and restore of configurations and data
- Open‑source with AGPL‑3.0 license
- Rich documentation and CLI/GUI management
- Supports custom knowledge bases and toolsets
- Supervisor daemon for uptime and auto‑restart
Disadvantages
- Requires an OpenAI‑compatible LLM endpoint (cost and availability)
- Learning curve for configuring agents, skills, and channels
- Docker needed for sandboxed execution (adds complexity)
- Remote execution via Evonet may raise security concerns
- AGPL license may limit commercial use
- Limited channel support (Slack planned)
- Heuristic safety checks may block legitimate actions
- Performance overhead from sandboxing and safety layers
- Dependency on external services (SSH, Docker, network)
- Documentation may not cover all edge cases
Installation
docker
docker build -t evonic-sandbox:latest docker/tools/
native
git clone https://github.com/anvie/evonic\ncd evonic\npython3 -m venv .venv\nsource .venv/bin/activate\npip install -r requirements.txt\nchmod +x ./evonic
FAQ
What is Evonic and what makes it unique compared to other agent frameworks?
Evonic is an agentic AI framework that lets you design, build, and orchestrate intelligent agents from concept to production. It stands out with three core differentiators: 1) **Workplace** – a first‑class execution environment that can be local, remote (SSH), or a tunnel via Evonet, enabling agents to run anywhere in your infrastructure; 2) **Agent‑to‑Agent Communication** – native messaging, delegation, and coordination between agents, making multi‑agent swarms and hierarchical orchestration a first‑class feature; 3) **Heuristic Mal‑activity Detection** – a real‑time safety system that inspects every action for dangerous patterns and escalates to a human operator instead of blindly executing.
How do I install and start Evonic on my machine?
You can install Evonic with a one‑liner: `curl -fsSL https://evonic.dev/install.sh | bash`. This clones the repo, sets up a virtual environment, installs dependencies, generates configuration, and adds `evonic` to your PATH. Alternatively, clone manually, create a venv, install requirements, and make the script executable. Once installed, start the server with `./evonic start` and open `http://localhost:8080` in a browser. If you want sandboxed tool execution, build the Docker image with `docker build -t evonic-sandbox:latest docker/tools/` and configure resource limits in `.env`.
How can I create, enable, and manage agents using the CLI?
Use the `agent` subcommands: `./evonic agent add my_bot --name "My Bot"` creates an agent; `./evonic agent enable my_bot` activates it; `./evonic agent disable my_bot` deactivates it; `./evonic agent remove my_bot` deletes it. You can list agents with `./evonic agent list` and view details with `./evonic agent get my_bot`. Agents are defined by six dimensions (concept, model, tools, knowledge base, channels, skills) and can be managed via the web UI under `/agents` as well.
What are Workplaces and how do I run agents on remote machines?
A Workplace is an execution environment abstraction. It can be a local sandboxed directory, a remote SSH server, or a tunnel using Evonet. To run an agent remotely, configure a remote Workplace in the agent’s settings or via the CLI, pointing to the SSH host or Evonet tunnel. This allows the agent to execute tools in that environment without needing a public IP or firewall changes.
How does Evonic keep agents safe from malicious or unintended actions?
Evonic implements a multi‑layer safety system: the **Injection Guard** blocks prompt injection and manipulation; the **Heuristic Mal‑activity Detection System** monitors every action for dangerous patterns like mass file deletion or unauthorized code execution; and the **Supervisor Daemon** watches agent health and restarts failures. Suspicious actions trigger escalation to a human operator, ensuring agents operate autonomously while maintaining security.
Featured in Videos
YouTube tutorials and walkthroughs for evonic
Alternatives
Similar projects ranked by category, topics, and text overlap.
