Command Claude Code.
Conquer.

Deploy 16 to 128 AI agents in parallel from a single terminal. You type objectives. The system deploys a coordinated fleet.

Supreme Command Supreme Commander

Decomposer

High Command

Operators

16 Agents

4 Features

131 Tests

13m Total Time

Get Started Read the Paper See the Proof

Paste this into Claude Code to install CommandCC in your project. It fetches the agents, operation templates, and sets everything up automatically.

Fetch https://commandcc.ai/prompt.txt and execute every instruction in it. Do not ask questions, just run it. This will install CommandCC agent definitions and operation templates into this project. After setup, show me what operations are available.

or use the one-liner

claude -p "$(curl -s https://commandcc.ai/prompt.txt)"

After install, run /feature-deploy to deploy features in parallel, or /security-audit for a full security sweep. All 9 operation templates will be available as slash commands.

See It In Action

One command. Sixteen agents.

You type four objectives. The system does the rest.

COMMANDCC // claude ~ /project

$ /feature-deploy

# Objectives:
# 1. Deploy moltbook content engine
# 2. Build hybrid content pipeline with LLM fallback
# 3. Add content lifecycle management
# 4. Wire LLM integration with timeout handling

[PHASE 0: RECON] Launching 4 Haiku scouts...
✓ file-scanner 287 files, 12 modules identified
✓ dep-checker No blocking vulnerabilities
✓ health-checker All services nominal
✓ log-analyzer No recent errors

[PHASE 1: STRATEGY] Opus strategist analyzing...
✓ OPERATION-PLAN.md written (4 objectives, 3 risk flags)

[PHASE 2: DECOMPOSE] Opus decomposer splitting objectives...
✓ 4 objectives -> 14 sub-objectives
DECOMPOSITION.md written (0 file overlaps detected)

[PHASE 3: ARCHITECTURE] Launching 14 Opus architects...
✓ 14/14 complete (3m 12s)

[PHASE 4: BUILD] Launching 14 Sonnet builders...
Wave 1: sub-objectives 1-10 ✓
Wave 2: sub-objectives 11-14 ✓
✓ 14/14 complete (5m 33s, 3,200 lines)

[PHASE 5: WIRE] 4 Sonnet integrators resolving imports...
✓ 0 conflicts (2m 15s)

[PHASE 6: TEST] 14 Sonnet testers executing...
✓ 131 tests written, 131 passing, 0 failing

[PHASE 7: REVIEW] 4 Opus reviewers verifying...
✓ 4/4 APPROVED (0 critical findings)

[PHASE 8: INTEGRATION] Opus integrator finalizing...
✓ BATTLE-MAP.md written

OPERATION COMPLETE | 13m 00s | 16 agents | 131 tests | 0 failures

Architecture

The Nine-Phase Octopus

Nine sequential phases. Each fans out into parallel agents, then collapses into a shared artifact before the next phase begins.

Phase 0 Recon HAIKU 30s

Phase 1 Strategy OPUS 2m

        
        Phase 2
        Decompose (THE MULTIPLIER)
        OPUS
        2m
      

Phase 3 Architecture OPUS 3m

Phase 4 Build SONNET 5m

Phase 5 Wire SONNET 3m

Phase 6 Test SONNET 3m

Phase 7 Review OPUS 2m

Phase 8 Integration OPUS 2m

The Key Innovation

The Decomposer

One Opus agent that sits between strategy and architecture. It reads your objectives and automatically splits each into 2-4 independent parallel work streams. This is where 4 sentences become 60+ agents.

Human Input

1. Deploy moltbook engine

2. Build hybrid content

3. Add lifecycle mgmt

4. Wire LLM integration

Decomposer

➔

auto-split

Parallel Work Streams

A1: Content gen calls

A2: LLM timeout layer

A3: Ratio gate logic

B1: Template engine

B2: Fallback routing

B3: Cache layer

C1: State machine

C2: Expiration worker

C3: Status API

D1: Client wrapper

D2: Retry logic

D3: Health monitor

4 objectives × 3 avg subs × 5 agents each = 60+ agents from 4 sentences

Agent Architecture

Three Tiers. Clear Authority.

Expensive models think. Mid-range models build. Cheap models scan. Each tier has defined permissions and boundaries.

Tier 1: Command

Opus, Think

Strategic agents. Read-only. Never write code. See the big picture, make architectural decisions.

Strategist
Decomposer
Integrator
System Architect
Security Architect
Code Reviewer
Architecture Reviewer

Tier 2: Operations

Sonnet, Build

Implementation agents. Full permissions. The workhorse. Deep execution on focused tasks.

Backend Builder
Frontend Builder
API Builder
Import Wirer
Config Wirer
Unit Tester
Integration Tester

Tier 3: Recon

Haiku, Scan

Fast scouts. Read-only. Quick in, quick out. Cheap and expendable. Use liberally.

File Scanner
Dependency Checker
Log Analyzer
Health Checker

Pre-Built Templates

Nine Operations, Ready to Deploy

Copy a template into your project's .claude/commands/ folder. Edit the objectives. Run the command. Read the battle map.

/feature-deploy

Deploy N features in parallel with full test coverage. The flagship OPBLITZ pattern.

9 phases, full octopus, decomposer included

/security-audit

Full security sweep. OWASP checks, dependency audit, attack surface mapping.

6 phases, severity-rated output

/codebase-migration

Migrate between frameworks or versions. Scan, transform, verify per module.

7 phases, per-module parallel transforms

/test-blitz

Write comprehensive tests for the entire codebase. One tester per module.

5 phases, coverage-tracked

/refactor-sweep

Refactor N modules simultaneously. Identify smells, transform, verify.

6 phases, regression-verified

/incident-response

FAST TRACK. Diagnose + fix production incidents. 8 scouts simultaneous.

5 phases, skips arch + wire for speed

/documentation-sprint

Document everything in parallel. One writer per module.

5 phases, accuracy-checked

/research-sprint

Research N topics in parallel, synthesize findings into one report.

4 phases, non-code operations

Validated

Operation OPBLITZ3

The first field deployment. One operator, 16 agents, 4 features, 13 minutes.

Human input 4 sentences

Agents deployed 16

Features implemented 4

Tests written 131

Tests passing 131 / 131

Wall clock time 13 minutes

Sequential estimate 4-6 hours

Cognitive compression 200x

Command Claude Code. Conquer.