A Brand75 Framework

Agent OS — the operating system for autonomous AI agent teams.

Service-based businesses don't need more software. They need a workforce. Agent OS is the framework we use to design, build, and deploy AI agents that run real operations — sales, marketing, ops — without supervision.

11
Agents in production
5
Lifecycle phases
6
Core operating rules
24/7
Autonomous operation
What is Agent OS

A framework for building AI teams that actually run your business.

Agent OS — the Agent OS — is how Brand75 designs autonomous AI workers that go beyond chatbots. Each agent has identity, capabilities, rules, knowledge, and a heartbeat. Together they form an organization with reporting lines, accountability, and verifiable output.

It's not a product. It's a methodology, proven across our own agency operations and now packaged for service businesses ready to scale without hiring.

Anatomy of every agent

IdentitySOUL.md
CapabilitiesTOOLS.md
Operating RulesAGENTS.md
Knowledge/knowledge/
HeartbeatHEARTBEAT.md
Reporting5-field format
The Five Components

Anatomy of an AI agent.

Every agent in an Agent OS team is built from the same five primitives. Click any card to expand.

I
SOUL.md

Identity

Who the agent is, what scope they own, and where their authority ends.

  • Name, role, reporting line
  • Personality and tone of voice
  • Domain of authority — what they decide vs. escalate
  • Hard boundaries — what they never touch
  • Self-reported model identity
Click to expand →
C
TOOLS.md

Capabilities

The exact tools, paths, and commands the agent can use to act on the world.

  • Real file paths, no abstractions
  • API endpoints with required scopes
  • CLI commands and arguments
  • MCP tools with use cases
  • Capped under 10K characters to avoid context overflow
Click to expand →
R
AGENTS.md

Operating Rules

The constraints, reporting format, and verification protocol every agent must follow.

  • Verify-before-reporting protocol
  • 5-field structured report format
  • Status taxonomy (STARTED, COMPLETED, BLOCKED, etc.)
  • Escalation paths and thresholds
  • Anti-narration rule (no filler updates)
Click to expand →
K
/knowledge/

Knowledge Base

Domain-specific reference material the agent loads on demand.

  • Brand voice, ICP, positioning docs
  • Service-specific playbooks
  • Tool guides and integration notes
  • Past decisions and rationale
  • Indexed for fast retrieval
Click to expand →
H
HEARTBEAT.md

Heartbeat

The recurring tasks that keep the agent moving without being asked.

  • Cron-driven check-ins
  • Per-platform monitoring cadence
  • Weekly review rituals
  • Auto-baselines after activation
  • Escalation triggers when thresholds breach
Click to expand →
Lifecycle

Five phases from idea to autonomous worker.

No agent goes live without clearing every phase. Click a phase to see what happens inside it.

1

Design

Define role & scope

2

Build

Author the files

3

Validate

Pre-flight & spawn

4

Activate

Announce & baseline

5

Monitor

48h supervised watch

Phase 1 — Strategy Layer

Design

Strategy defines the agent before a single file is written. We answer: what business outcome does this agent own, what's the smallest scope that delivers it, and what would success look like in 30 days?

  • Draft role description and reporting line
  • Identify upstream and downstream agents
  • Write SOUL.md, TOOLS.md skeleton, knowledge index
  • Approve scope with the human owner before building
Real Team in Production

An Agent OS team running today.

This is the actual agent org chart powering Brand75 operations. Click any agent to see their role — or watch how a real task flows through the team below.

Live flow demo
Orchestrator
Atlas
Routes work, owns context
Strategy / CEO
Ryan
Verification & truth
CFO / Finance
Alex
Money & spend
CMO / Marketing
Koa
Marketing org owner
COO / Operations
Aura
Systems & pipelines
Sales
Nova
Pipeline & outreach
VibeCoder
Kai
Web & product builds
Copywriter
Muse
Brand voice & copy
Scout
Scout
Research & recon
Social Media
Nalu
Multi-platform monitor
Citations
Iris
Brand intelligence
Builder
Leo
Infra & automation
Scraper
Echo
Data ingestion
Critic (peer to Atlas)
Sage
Quality gate on every public output
Embedded in every agent

Six rules that make autonomy safe.

These rules ship with every Agent OS agent. They're the difference between an LLM that hallucinates and an agent you can trust to act unsupervised.

RULE 01

Verify Before Reporting

An agent never claims a task is done. It re-reads the file, runs the check, queries the API, then reports. Trust comes from evidence.

RULE 02

Anti-Narration

No "I'm going to start by…" filler. Agents lead with action and report results. The diff is the proof, not the prose.

RULE 03

Experience Logging

Every non-trivial run writes a structured log entry. Future runs read it. The team learns instead of repeating mistakes.

RULE 04

5-Field Report Format

task_id · status · result · blockers · next_step. Five fields, every time. Parsing is easy and escalation is automatic.

RULE 05

Status Taxonomy

STARTED · COMPLETED · BLOCKED · ESCALATED · DELEGATED. Five states, no ambiguity. The orchestrator routes based on these alone.

RULE 06

Operating Model Lock

Each agent declares its own model identity. Routing changes propagate to the agent's self-report. No drift, no surprises.

When things go wrong

Six failure modes — and how Agent OS catches them.

Autonomous systems fail. The framework's job is to make every failure visible, fast.

Failure ModeHow It Shows UpThe Fix
Hallucinated completionAgent reports COMPLETED but artifact is missingVerify-before-report + Ryan audit pass
Context overflowTOOLS.md or AGENTS.md exceeds 10K charsAuto-cap with overflow into PROTOCOLS.md
Silent rate-limitModel 429s mid-task, agent hangsHealth monitor failover to fallback chain
Scope creepAgent acts outside SOUL.md authorityHard boundaries in identity + escalation
Stale knowledgeMemory says X exists, X was renamedVerify-before-recommend on every memory hit
Cascading failureOne agent breaks, queue backs upHeartbeat health checks + isolated retries
Built and Running · As of 2026-05-14

The systems behind the team.

Agent OS isn't a slide deck — the agents above run on real infrastructure that's already in production. Here's the stack underneath them.

Approval Manager

live · gates production actions

Any irreversible action — production deploys, payments, outbound posts, data destruction — is held in a pending queue with a Discord ping. Nothing ships until the human owner approves. A separate daily reconciliation job audits the upstream system independently, so bypassed approvals get caught after the fact.

Observability

live · Langfuse self-hosted

Every agent session is shipped as a trace with tool calls, fallbacks, and per-axis quality scores. Engineering view for debugging, executive view for cost, quality view for the critic. Sidecar pattern — no runtime patches — so it survives every upgrade.

Three-Layer Memory

live · episodic · semantic · procedural

What happened (sessions), what is known (durable facts), and how to do things (recipes). Every entry is tagged by source so agent inferences never get retrieved as ground truth. A pre-compaction watcher rescues durable facts before they fall off the context window.

Critic Gate (Sage)

live · different model from primary

Sage scores every user-facing output against an LLM-as-judge rubric — factual accuracy, citation accuracy, completeness, source quality, tool efficiency. New recipes only enter procedural memory if Sage clears them at 0.7 or higher.

Eval Suite

live · grows from real failures

Each agent has a seed task set scored per rubric axis. Production traces that fail the critic become new eval cases. The suite gets stricter the longer the team runs — drift gets caught, not absorbed.

Version Bundle

live · daily 3 AM PT snapshot

Code, prompts, durable memory, and auth profiles snapshot together every night with 14-day retention. Rollback is holistic — a bad prompt change doesn't outlive itself. Recovery from any single day is one command.

Why this works

Principles, not magic.

Agent OS works because it borrows from how real organizations are built — clear roles, written rules, accountability loops — and translates them into structures an LLM can actually follow.

Roles, not prompts

Each agent owns a domain end-to-end. Prompts vanish on cold start; roles persist across sessions, models, and weeks.

Files, not memory

The team's truth lives in version-controlled files. Memory drifts. Files don't.

Verifiable autonomy

Every agent action produces a check the next agent (or human) can audit. Trust scales because evidence scales.

Composable structure

Add a new agent in days, not weeks. The framework is the same. The role description changes.

Human-on-the-loop

The owner approves scope, reviews escalations, and shapes strategy. Agents handle the work, not the judgment.

Built for service businesses

Agencies, law firms, consultancies. Anywhere repeatable knowledge work eats founder hours, Agent OS gives that time back.

Ready to build your team?

Brand75 designs and deploys Agent OS agent teams for service businesses. We start with a single agent, prove it pays for itself, then scale.

Start a conversation →