Open Source · Apache 2.0 · Kubernetes-Native

Build AI Agent Teams That Run Your Business Processes

Declare AI agents as Kubernetes CRDs, wire them into DAG workflows, and connect them to any LLM. From marketing campaigns to code reviews to contract analysis — Purko handles it.

$ helm install purko oci://ghcr.io/purko-io/purko
Get Started Documentation →
Purko Controller triage anomaly remediate capacity PagerDuty Prometheus Slack GitHub

The Problem

Your Team Is Stuck on Repetitive Work

3–5 days
Manual Processes

for every campaign, contract review, or report. AI can do each task, but someone still has to run the process — copy outputs, check quality, hand off to the next step.

All or Nothing
Trust Problem

AI tools either run free or need constant oversight. No middle ground. Teams don't trust AI to handle client-facing work, so they supervise everything — eliminating the productivity gains.

One-Shot
No Process Automation

ChatGPT handles single prompts. Your business runs on processes — sequences of tasks that require multiple specialists, quality checks, and handoffs. No AI tool orchestrates end-to-end.

Your team wastes hours on tasks AI could handle in minutes. Purko changes that.

The Solution

Declare Your Agents. Kubernetes Runs Them.

  • ▶ Agents are Kubernetes resources — deploy with kubectl apply, manage with GitOps, govern with RBAC. Same workflow as everything else in your cluster.
  • ▶ MCP connectors unify your tools — GitHub, Slack, Jira, and any MCP-compatible tool server feed into a single agent context. Connect your existing systems in minutes.
  • ▶ DAG-based workflows orchestrate multi-agent operations — fan-out/fan-in, conditional branching, retry policies, and cross-agent data passing. Define once, execute reliably.
  • ▶ Open-source core (Apache 2.0) — inspect the code, fork it, contribute back. No black boxes in production. Enterprise features under BSL convert to Apache 2.0 after 36 months.
agent-triage.yaml 
apiVersion: purko.io/v1alpha1
kind: Agent
metadata:
  name: incident-triage-prod
  namespace: sre-agents
spec:
  agentType: incident-triage
  autonomyLevel: "ha"
  model:
    provider: anthropic
    name: claude-sonnet-4-20250514
  guardrails:
    maxActionsPerMinute: 20
    requireApproval: false
    maxBlastRadius: "pod"
    rollbackOnFailure: true

Graduated Autonomy

Trust Is Earned, Not Configured

Purko agents graduate from supervised to autonomous through the Shu-Ha-Ri model — inspired by martial arts mastery. Promotion is earned through measurable reliability.

守
Shu
Follow
Manual Approval

Agent strictly follows predefined runbooks. All actions require explicit human approval before execution.

< 100 actions · Building trust
→
破
Ha
Adapt
Semi-Autonomous

Agent adapts procedures based on context. Low-risk actions auto-approved; high-risk still requires human confirmation.

95%+ success · 14+ days proven
→
離
Ri
Transcend
Fully Autonomous

Agent operates autonomously within defined safety boundaries. Full self-healing with policy-gated execution.

99%+ success · 500+ actions · Battle-tested
incident-triage-prod
247 actions 97% success 3% override rate
Shu ▶ Ha (current) Ri

Agent Catalog

Pre-Built Agents for Any Business

34 production-ready agents across 5 categories. Each one is a CRD — version-controlled, testable, observable.

✎
Ready

Content Writer

Produces reports, proposals, emails, blog posts, and social content adapted to your audience and brand voice.

supervised
◎
Ready

Quality Reviewer

Checks content, contracts, and code for accuracy, compliance, and quality. Returns structured verdicts with specific fix suggestions.

restricted
◆
Ready

Data Analyst

Analyzes data, identifies patterns and anomalies, and produces executive-ready insights with specific recommendations.

restricted
▣
Ready

Research Analyst

Investigates topics using available tools, synthesizes findings from multiple sources, and assesses confidence levels.

restricted
⚙
Ready

Code Executor

Writes and executes code in a sandboxed environment. Handles Python, bash, data processing, and automated testing.

supervised
↻
Ready

Project Coordinator

Manages timelines, tracks deliverables, flags risks, and ensures handoffs between team members happen on schedule.

full

Any Industry

One Platform. Every Business Process.

Purko's six agent types — planner, executor, reviewer, router, monitor, and retriever — map to every business function. Marketing campaigns, legal contract review, real estate listings, data analytics, video production, software development.

The same CRDs, the same DAG workflows, the same trust model. Define any business process as a Workflow and Kubernetes orchestrates it — from intake through delivery.

Build your AI team in minutes. Each agent is a CRD with a system prompt, model config, and tool bindings. Need a contract analyst? A content writer? A campaign strategist? Define the spec, apply it, done.

Purko works for marketing agencies, law firms, real estate brokerages, data consultancies, video studios, and software teams — any process where multiple AI specialists need to collaborate.

▣

Plan

planner agent
→
✎

Code

executor agent
→
◎

Review

reviewer agent
→
⚙

Test & Build

executor agent
→
▶

Deploy

router agent
→
◆

Monitor

monitor agent
workflow-dev-pipeline.yaml 
apiVersion: purko.io/v1alpha1
kind: Workflow
metadata:
  name: code-development-pipeline
spec:
  timeout: "45m"
  onFailure: stop
  steps:
    - name: plan
      agentRef:
        name: project-planner
    - name: implement
      agentRef:
        name: code-executor
      dependsOn: [plan]
    - name: review
      agentRef:
        name: code-reviewer
      dependsOn: [implement]
agent-code-reviewer.yaml 
apiVersion: purko.io/v1alpha1
kind: Agent
metadata:
  name: code-reviewer
spec:
  type: reviewer
  model:
    name: claude-sonnet-4-20250514
    provider: anthropic
    temperature: 0.1
  tools:
    - name: static-analysis
      type: builtin
    - name: vulnerability-scan
      type: api
  guardrails:
    maxIterations: 5
    costLimitUSD: 3.0

Integrations

Connects to Everything You Already Use

MCP-native connectors unify your tool stack into a single agent context. No custom integrations required.

PagerDuty
Slack
Prometheus
Datadog
Jira
GitHub
Grafana
OPA / Kyverno
Splunk
ServiceNow
New Relic
Terraform

Community-maintained connectors are open source. Enterprise connectors available with Pro and Enterprise tiers.

Get Started

Build Your AI Team In Minutes

Purko is open source and ready to use. Install with Helm, create your first agent, and run your first workflow.

$ helm install purko oci://ghcr.io/purko-io/purko
View on GitHub Read the Docs →
Open Source · Apache 2.0 · 26 CLI Commands · 34 Pre-Built Agents