Open Source Lab

AgenticBusiness OS

A practical, open-source workspace template for running a business with AI agents, reusable workflows, approvals, memory, schedules, and source-backed outputs.

What It Is

An operating workspace for AI-powered work

Most agent projects stall because context lives in chats and workflows live in someone's head. Agentic Business OS gives the assistant a structured place to understand the business, run repeatable work, and keep humans in control.

Skills and Agents

Define repeatable jobs as skills, then use agent roles for focused work like analysis, support triage, growth ops, QA, and automation building.

Repeatable JobsRole ContractsClear Ownership

Approval Boundaries

Let agents prepare work, gather evidence, and recommend actions while humans approve external writes, customer-facing messages, publishing, and sensitive changes.

Review FirstControlsEscalation

Business Memory

Keep durable context in one place so agents can reuse decisions, source cards, operating rules, previous outputs, and hard-won lessons.

ContextSourcesHistory

Operating Rhythm

Create daily, weekly, and event-triggered routines so the business gets a steady stream of useful agent work instead of random AI experiments.

SchedulesReportsChecks
How It Works

Simple layers, real operating discipline

The website should explain the concept. The GitHub repo is where the setup docs, templates, privacy rules, and implementation details belong.

Project context

Start with goals, tools, routines, risks, operating rules, and the first workflows worth making repeatable.

Skills and runbooks

Turn repeated work into explicit procedures the assistant can follow, improve, and hand back with evidence.

Memory spine

Route generated material through inbox, durable synthesis through wiki, source material through sources, and final work through outputs.

Schedules and reviews

Run daily, weekly, and recurring checks so important operational work does not depend on someone remembering to ask.

Human approval gates

Keep risky actions review-first: publishing, messaging, external updates, credentials, billing, and destructive changes.

Outputs with evidence

Produce briefs, reports, task handoffs, decisions, and logs that can be inspected, reused, and improved over time.

Used In The Wild

Tested inside Caliverse

We use this system in our own fitness app business, Caliverse. That matters because it has to survive real operating work: analytics, growth, support, product, finance, tooling, and planning.

Caliverse
Dogfood workspace
7
Operating domains
Skills
Workflow model
Review-first
Safety model

Daily operating rhythm

Use plans, recurring reviews, scheduled checks, and memory ingest so work compounds instead of disappearing into chat history.

Analytics and growth signals

Turn analytics, campaign, search, review, and app-store signals into reports, decisions, and follow-up work.

Support and product handoffs

Summarize support themes, prepare public replies for approval, and convert repeated product signals into actionable handoffs.

What Ships

A starter workspace, not magic dust

The repository is meant to be the implementation and documentation home. It gives operators and builders a privacy-safe template instead of hiding the system behind a black box.

agentic-business-os
README.md
AGENTS.md
00-start-here.md
context/README.md
domains/README.md
inbox/README.md
wiki/README.md
sources/README.md
outputs/README.md
.agents/skills/ops/project-onboarding/
system/tools/template_sync.py
docs/privacy.md
docs/onboarding.md
docs/sync.md
evals/README.md

Workspace template

Starter files for project context, operating domains, rules, source routing, durable memory, and project navigation.

ContextDomainsMemory Spine

Skill scaffolds

Patterns for turning repeated work into explicit procedures with ownership, inputs, related agents, approvals, and expected outputs.

SkillsAgentsSchedules

Sync and tooling

Deterministic tools for audits, template sync, onboarding packs, evals, and safe export from a private workspace to the public template.

AuditsTemplate SyncEvals

Docs stay in GitHub

Setup, privacy, onboarding, sync, folder contracts, and implementation guidance live with the open-source project.

SetupPrivacyFolder Contracts
Where We Help

Try it yourself. Bring us in when it gets real.

Open source should let people learn and experiment. Our commercial role is helping you turn that experiment into a working business system.

Setup and customization

We help adapt the OS to your tools, workflows, approval rules, and operating rhythm.

Custom agent builds

We build the first high-value agents with you so the system starts with real business impact.

Support when stuck

Bring us in when integrations, permissions, reliability, or workflow design become more complicated than expected.

Build Your AI Operating Layer

Start with the open-source system, or ask us to help shape it around your business from day one.