quarkloop
Use CasesPricingDocsBlog
Book a DemoGet Started
quarkloop

Your Agents. Your Machine. Your Control.

Product

  • Use Cases
  • Enterprise
  • Pricing
  • Integrations

Resources

  • Documentation
  • Getting Started
  • Quarkfile
  • Blog

Company

  • About
  • Contact
  • GitHub

© 2026 Quarkloop. Apache 2.0 License.

PrivacyTerms
Open SourceApache 2.0v2026.04Enterprise

Enterprise-grade multi-agent AI.
Structured autonomy at scale.

Autonomous AI Agents for Everyday Work — Your Agents. Your Machine. Your Control.

Get started — freeBook a demo
Terminal
$ quark space run  Loading Quarkfile ...  Initializing plugins: medical-terminology, hipaa-compliance  Starting inboxes: patient-forms, insurance-claims  Main Agent ready  Channels active: web:3000, telegram   [main-agent]  ORIENT   reading inbox 'patient-forms' ...  [main-agent]  PLAN     spawning sub-agent: triage-bot  [triage-bot]  EXECUTE  processing intake form  [triage-bot]  SESSION  user sess_abc123 asking for status  [triage-bot]  COMPLETE exit code 0
Apache 2.0License
4Core Components
3Execution Modes
7Industries
LocalRuntime

Why Quark

Built for enterprise autonomy.

The architecture that makes Quark different from opaque, cloud-hosted agent platforms.

Your Control

Agents run on your machine. Every action is auditable and reversible. No cloud dependency, no data leaving your infrastructure.

Three execution modes: Autonomous, Assistive (human-in-the-loop), and Workflow (deterministic DAG).

Infrastructure as Code

One Quarkfile defines your entire Space: agents, plugins, inboxes, channels. Version it, share it, reproduce it anywhere.

Like a Dockerfile for AI agents. Space = built environment. Runtime = execution engine.

Omnichannel Collaboration

Channels connect users to agents through Web, Telegram, WhatsApp. Sessions maintain isolated conversational context.

Session Context never pollutes Work Context. Human queries don't disrupt agent processing.

Scalable Data Ingestion

Inboxes are async message buses for external data. Sub-agents consume messages as they arrive. No polling, no bottlenecks.

Main Agent creates Inboxes. Sub-agents subscribe and process at scale.

How it works

Three commands to autonomous agents.

From zero to a running multi-agent Space in under a minute.

$ quark space init healthcare-ops  Created Quarkfile  Created plugins/  Created prompts/
1

Define your Space

A Space is your workspace + execution environment. Like a Docker image for agents — define once, run anywhere.

$ quark space run  Loading Quarkfile ...  Initializing plugins: medical-terminology  Starting inboxes: patient-forms  Main Agent ready  Channels active: web:3000
2

Start the Runtime

The Runtime launches your Space. Plugins load, Inboxes open, Channels connect. The Main Agent begins orchestrating.

$ quark activity  [main-agent]  ORIENT   reading inbox ...  [main-agent]  PLAN     spawning triage-bot  [triage-bot]  SESSION  user asking for status  [triage-bot]  COMPLETE exit code 0
3

Watch Agents Collaborate

The Main Agent orchestrates. Sub-agents handle specific tasks. Sessions maintain conversational context.

Under the hood

The Main Agent Loop

An OODA-inspired continuous cycle. The Main Agent orients on Inbox state, plans next steps, spawns Sub-agents, monitors execution, then assesses outcomes — and repeats.

OrientRead KB state
PlanGenerate steps
DispatchSpawn agents
MonitorTrack progress
AssessEvaluate results
Continuous cycle

Configuration

One file. Single space. Autonomous Agents.

Define your agents, tools, and model in a Quarkfile.
Run quark run to start.

Quarkfile
quark: "2026.04" meta:  name: healthcare-intake  version: 1.0.0 space:  type: healthcare  execution_mode: assistive model:  provider: anthropic  name: claude-sonnet-4-20250514 plugins:  skills:    - ref: quark/medical-terminology    - ref: quark/hipaa-compliance  binaries:    - ref: quark/pdf-parser main_agent:  prompt: ./prompts/main-agent.md  sub_agents:    - name: triage-bot      prompt: ./prompts/triage.md      consumes: [patient-forms]    - name: billing-reconciler      prompt: ./prompts/billing.md      plugins: [erp_integration] inboxes:  - name: patient-forms    type: queue    retention: 7d  - name: insurance-claims    type: queue channels:  - type: web    port: 3000  - type: telegram    token: $TELEGRAM_BOT_TOKEN memory:  episodic: ./state/work-context.db  semantic: ./state/vector-store.db env:  - ANTHROPIC_API_KEY  - TELEGRAM_BOT_TOKEN
Terminal
$ quark space init healthcare-intake  Created Quarkfile  Created prompts/  Created plugins/ $ quark space run  Loading Quarkfile ...  Initializing plugins: medical-terminology, hipaa-compliance  Starting inboxes: patient-forms, insurance-claims  Main Agent ready  Channels active: web:3000, telegram $ quark activity  [main-agent]  ORIENT   reading inbox 'patient-forms' ...  [main-agent]  PLAN     spawning sub-agent: triage-bot  [triage-bot]  EXECUTE  processing intake form  [triage-bot]  SESSION  user sess_abc123 asking for status  [triage-bot]  COMPLETE exit code 0

Comparison

Quark vs. the cloud.

Local execution vs opaque cloud platforms. Your agents, your machine, your control.

Feature
Quark
Cloud Platforms
Execution model
Quark:Local & auditable
Cloud:Opaque
Agent hierarchy
Quark:Main + Sub-agents
Cloud:Flat / proprietary
Human-in-the-loop
Quark:Native (Assistive mode)
Cloud:Add-on / limited
Data ingestion
Quark:Async Inboxes
Cloud:API polling
Session management
Quark:Built-in Channels
Cloud:External integrations
Memory architecture
Quark:Episodic + Semantic
Cloud:Varies
Plugin system
Quark:Skills + Binaries
Cloud:Proprietary SDKs
Configuration
Quark:Quarkfile (YAML)
Cloud:Dashboard / API
License
Quark:Apache 2.0
Cloud:Proprietary

Open source core

Built in the open.
Trusted in production.

The best infrastructure is transparent. Quark's open-source core gives you full visibility and control — no vendor lock-in, no surprises.

Star on GitHubContributing guide

Apache 2.0 licensed

Use, modify, and distribute. Supervisor, Runtime, CLI, and Web Channel — all open source. No feature gating.

Community-driven

Modular Golang architecture. Four core components, each independently maintainable and extensible.

No telemetry, no tracking

Agent context, session history, and semantic memory stay on your infrastructure. Fully auditable execution.

Foundation for Enterprise

Quark Enterprise adds SSO, audit logs, and compliance features on the same open core you trust.

Quark Enterprise

Enterprise features.
Same local runtime.

Quark Enterprise adds the features teams need to run autonomous agents in production — without changing how the runtime works.

Learn about Enterprise

SSO & LDAP Integration

Connect Quark to your identity provider. Authenticate users and control agent Space access through your existing SSO or LDAP directory.

Priority Support with SLA

Guaranteed response times with a dedicated support channel. Direct access to the engineering team. Escalation paths for production incidents.

Private Deployment Assistance

Deploy Quark in air-gapped networks with custom infrastructure. We help you run Spaces securely in your environment.

Get started

Ready to deploy autonomous
agents with confidence?

Start with open-source Quark. Define Spaces, configure Inboxes, connect Channels. When you need compliance and enterprise support — we're one conversation away.

Get started — freeBook a demo