Launch App
LMC PRO ENGINE V2.4 IS LIVE

Less prompt.
More build.

Deploy a coordinated cohort of specialized AI agents. Synchronize local workspaces, run consensus-driven code synthesis, and automate software delivery.

$npm install -g @giant-agent/lmc
Documentation → GitHub Repository →

Why Giant Agent LMC?

Single-agent prompting produces code snippets. LMC coordinates specialized cohorts to write full-stack production software systems.

Decentralized Consensus

Agents vote on design patterns, architectural choices, and library selections before writing a single line of code.

State Sync Engine

Dynamic local folder sync that reflects workspace adjustments, environment alterations, and codebase changes live.

Keyboard-First TUI

Control running agent cohorts directly from your terminal using shortcuts. No heavy GUI dependencies required.

Live Workspace Diff

Watch specialized agents modify scripts, write test files, and perform code refactoring live in standard Git-diff styles.

Goal Decomposer

Automatically breakdowns ambiguous user prompts into structured modular sub-tasks, allocating them to specialized agents.

Agent-to-Agent Mesh

Direct low-latency communication layer using specific JSON messaging schemas, bypassing slow central queues.

Sandbox Execution

Runs self-contained workspace environments safely. Installs required libraries, executes code, and audits outputs.

Cost Optimization

Dynamically matches task complexity with the cheapest capable models in real-time, reducing LLM token bills by 60%.

Swarm Dashboard Terminal

Click the TUI tabs below to explore how the multi-agent coordinate mesh operates in real-time.

giant@lmc-node-01:~
ACTIVE SWARM
  • ● coordinator
  • ● arch-agent
  • ● code-gen-01
  • ● security-auditor
WORKSPACE FILES
  • ✎ server.js
  • ✚ auth.test.js
  • 📄 config.json
  • 📄 index.html
  • ✕ legacy.js
=================== SWARM ORCHESTRATION TASK MAP =================== Goal: Build Secure JWT Authentication Module with Unit Tests └─ [x] Decompose prompt into architectural goals ├─ [x] Setup environment sandboxing [DONE] ├─ [x] Design DB schemas (users table) [DONE] ├─ [/] Generate secure auth backend module [IN PROGRESS] │ ├─ [x] Write helper functions (bcrypt hash) [DONE] │ ├─ [/] Implement JWT signing middleware [COMPILING] │ └─ [ ] Set up login rate limiting [PENDING] ├─ [ ] Create Mocha/Chai test scripts [WAITING] │ └─ [ ] Verify negative testing scenarios [WAITING] └─ [ ] Conduct static analysis & vulnerability audit [QUEUED] ──────────────────────────────────────────────────────────────────── Cohort: Local Sandboxed Engine | Models: 4x Gemini-3.5-Flash
PROPOSAL #104: Use bcryptjs over native crypto-scrypt PASSED (3/4)
● arch-agent (YES) ● code-gen-01 (YES) ● security-auditor (NO: preferred native) ● coordinator (YES)
PROPOSAL #105: JWT expiration policy setting (15m vs 24h) VOTING IN PROGRESS
● arch-agent (YES: 15m) ● code-gen-01 (NO: prefers 24h for development ease) ● security-auditor (YES: 15m is audit standard) ● coordinator (UNDECIDED)
diff --git a/server.js b/server.js Active Editor: code-gen-01
12 const jwt = require('jsonwebtoken');
13 const bcrypt = require('bcryptjs');
14-function generateToken(userId) {
15- return jwt.sign({ id: userId }, 'secret_key');
16-}
14+function generateToken(userId) {
15+ // Updated securely: Fetch secret from environment and set expiration to 15m
16+ return jwt.sign({ id: userId }, process.env.JWT_ACCESS_SECRET, {
17+ expiresIn: '15m'
18+ });
19+}
20
21 module.exports = { generateToken };
code-gen-01 ➔ security-auditor 01:31:02
{"type": "AUDIT_REQUEST", "file": "server.js", "lines": "14-19"}
security-auditor ➔ code-gen-01 01:31:04
{"status": "PASS", "comments": "Bcrypt salt rounds (10) and JWT expiration verified.", "risk_level": "LOW"}
coordinator ➔ workspace 01:31:05
{"command": "COMMIT_STATE", "files": ["server.js"], "message": "feat: enhance JWT signing security"}
LMC Prompt Shell initialized. Type 'help' to see active commands.
status
Cohort is active. 4/4 agents spawned, sandbox secured. Current network cost: $0.0042. Ready.
>

Swarm Compiler Studio

Configure custom agent parameters, select specialized roles, and initialize a collaborative build session live.

Swarm Setup LMC-Local
Autonomy Level High
Consensus Threshold 75%
Specialized Agent Cohort
Architect
Sets patterns & plans task hierarchies
Coder
Writes modules & refactors state
Auditor
Checks vulnerabilities & code smells
DevOps
Compiles files & builds docker swarms
Live Swarm Log Output
💻

Swarm Idle

Adjust parameters on the left and click "Initialize Swarm Collaboration" to spin up the local model workspaces.

Giant Agent Manual Reference

Technical reference guides for managing Large Model Collaborators (LMC) via standard command lines.

1. Quickstart CLI Guide

Get started by executing LMC in any local workspace folder. Initialize a sandboxed swarm immediately:

$ lmc init --swarm="local" --models="gemini-flash"

This creates a hidden configuration file .lmc/config.json in your local folder defining the specialist agents, permissions, sandboxing guidelines, and maximum budget tokens.

2. Launching Collaborative Goals

Specify any target objective using standard prompt language. LMC automatically parses, allocates, and coordinates execution:

$ lmc run "Write secure JWT signup backend and create unit tests"

Key parameters you can pass to fine-tune operations:

  • --autonomy=[1|2|3]: Configures model self-execution limits (Low, Medium, High).
  • --consensus=[50..100]: Agreement margins required before code merges.
  • --sandbox=[docker|local]: Defines environment encapsulation.
3. API State Synchronization

LMC exposes local filesystems via P2P web socket streams. Integrate Swarms into existing Next.js or React build engines with our native API:

const { SwarmManager } = require('@giant-agent/lmc');

const swarm = new SwarmManager({
  configPath: './.lmc/config.json',
  autonomy: 'extreme'
});

// Stream state events live
swarm.on('workspace:diff', (diff) => {
  console.log(`[Coder Agent] File: ${diff.file} updated.`);
});

await swarm.deploy();

Transparent Swarm Licenses

Choose the model bandwidth, concurrency limits, and agent limits suitable for your workflow complexity.

Developer Cohort
$0/mo
  • Up to 3 concurrent agents
  • Local sandboxing engine
  • Git and directory sync
  • Standard P2P Agent Mesh
  • Dynamic model consensus
  • Docker sandbox isolation
Get Started
Consortium Swarm
$49/mo
  • Up to 10 concurrent agents
  • High Autonomy Level execution
  • Dynamic model voting consensus
  • Advanced cost optimization routing
  • Docker container sandbox isolation
  • Enterprise TUI server sync
Upgrade Swarm
Enterprise Hive
$299/mo
  • Unlimited concurrent agents
  • Full multi-node P2P clusters
  • Custom specialist fine-tuning
  • Real-time human-in-the-loop triggers
  • Dedicated virtual private sandboxes
  • Premium TUI server cloud dashboards
Contact Sales