Warning
Jackadi is currently in an alpha state and is provided with no guarantees or warranties. You're welcome to try it out and give feedback.
But, it must not be used in production or in any environment where reliability, security, or data integrity are required.
By using Jackadi at this stage, you acknowledge the risks and potential for breaking changes.
Jackadi is a developer-first distributed task execution platform designed for developers with a plugin system architecture consisting of a manager and agents.
The main motivation is to create a framework where developers write tasks as pure code without abstractions or hidden behaviors. Task writing is meant to be natural and direct.
Key principles:
- Pure Go Approach: Tasks are written as Go code with no hidden behaviors - what you write is what you get.
- No Runtime Dependencies: Tasks have no runtime dependencies on other tasks; all dependencies are resolved at compile-time.
- No Abstractions: Task writing is natural for Go developers with minimal framework-specific knowledge needed.
- Flexible Use Cases: From simple package installation to complex workflows like server management and upgrades.
| Feature | Description |
|---|---|
| Distributed Task Execution | Execute tasks across multiple agents from a central manager. |
| Plugin System | Extend functionality through custom Go plugins. |
| Advanced Targeting | Target agents via list, glob, regex, advanced query. |
| Specs Collection | Gather and store system information from agents. |
| Security | mTLS, agent acceptance workflow, protection against rogue agents. |
| Developer-Friendly | Tasks/specs are Go function registered with a simple SDK. |
| Web API | Integrate Jackadi with your infrastructure stack. |
Full documentation can be found here.
In a nutshell:
- Agents are connected to a manager via persistent bidirectional gRPC.
- Simple plugin system:
- All tasks and specs collectors are pure Go functions.
- The plugin system is based on hashicorp/go-plugin.
- The SDK is simple and easy to use.
- Tasks results are stored in a local BadgerDB.
# Start the manager
manager --mtls=false
# Start an agent
agent --id="agent1" --mtls=false
# Accept the agent connection (if not using auto-accept)
jack agents list
jack agents accept agent1
# The agent should be now in "accepted" list
jack agents list
# Check agents health
jack agents health
# Run a task
jack run agent1 cmd:run "echo hello"package main
import "github.com/jackadi-io/jackadi/sdk"
func Hello(name string) (string, error) {
return fmt.Sprintf("Hello %s!", name), nil
}
func main() {
tour := sdk.New("tour")
tour.MustRegisterTask("hello", Hello).WithDescription("Greetings.")
sdk.MustServe(tour)
}CGO_ENABLED=0 go build -o tour .Copy the file in the manager /opt/jackadi/plugins directory.
Then configure /etc/jackadi/plugins.yaml file:
"*":
- tourjack run agent1 plugins:syncjack run agent1 tour:hello