MoltBot Docs

Everything you need to deploy, manage, and scale autonomous AI agent fleets on dedicated GPU infrastructure.

⚡

Quickstart

Deploy your first agent in 5 minutes
🌐

Swarm Guide

Orchestrate multi-agent pipelines
📡

API Reference

Full REST API documentation
🤖

Model Routing

7 AI models, auto-optimized

Quickstart

Get a MoltBot agent running in under 5 minutes. You'll need a MoltBot Cloud account and an API key.

1. Install the MoltBot SDK

npm install @moltbot/sdk
# or
pip install moltbot

2. Initialize your client

import MoltBot from '@moltbot/sdk';

const client = new MoltBot({
  apiKey: process.env.MOLTBOT_API_KEY,
  region: 'us-east-1',
});

3. Deploy your first agent

const agent = await client.agents.deploy({
  name: 'my-first-agent',
  model: 'auto',        // Smart model routing
  tier: 'base',         // RTX 4090, $49/mo
  task: 'code-review',
});

console.log(`Agent deployed: ${agent.id}`);
// → Agent deployed: mb-a7x9f2

4. Run a task

const result = await agent.run({
  prompt: 'Review this PR and identify bugs',
  context: { pr_url: 'https://github.com/...' },
});

console.log(result.output);

Core Concepts

🤖 Agents

An agent is a persistent process running on a dedicated GPU VM. Each agent has its own memory, tool access, and model configuration. Agents run 24/7 and can be given recurring tasks or respond to triggers.

🌐 Swarms

A swarm is a coordinated group of agents working on a shared objective. MoltBot's orchestrator automatically assigns subtasks, manages dependencies, and aggregates results. Available on Swarm Matrix tier.

🧠 Memory

Each agent has persistent vector memory powered by ChromaDB. Memory survives reboots and can be shared across agents in the same organization. Semantic search enables agents to recall relevant context from past runs.

⚡ Model Routing Auto

MoltBot's arbitrage engine automatically selects the cheapest model capable of completing your task. Set model: 'auto' to save up to 90% on inference costs without sacrificing quality.

Supported Models

Model Provider Context Best For
Claude Opus 4.6Anthropic200KComplex reasoning
GPT-5.4OpenAI128KCode generation
Gemini 2.5 ProGoogle1MLong context analysis
Qwen 3.5 FreeNVIDIA NIM32KFast, cost-free routing
Kimi k2Moonshot128KMultilingual tasks
Grok 3xAI131KReal-time search
DeepSeek R2DeepSeek128KMathematics, STEM

GPU Tiers

Base — RTX 4090 (24GB VRAM) · $49/mo

1 persistent agent. Ideal for code review, content generation, data analysis, and single-threaded automation.

  • 24GB VRAM, 128GB system RAM
  • Qwen 3.5 local + cloud model routing
  • 10GB persistent storage

Swarm Matrix — A100 80GB · $149/mo

3 concurrent agents with full model access. Smart routing across all 7 providers. Best for teams and complex pipelines.

  • 80GB VRAM, 256GB RAM
  • All 7 AI models
  • 50GB persistent storage + ChromaDB memory

Enterprise — 4× H100 Cluster · $299/mo

10 concurrent agents, Omnisphere pipeline builder, priority support, and custom SLA.

  • 4× H100 80GB cluster, 512GB RAM
  • 200GB storage + dedicated vector DB
  • Visual pipeline builder, webhook triggers, custom models

Plans & Billing

MoltBot uses monthly subscription billing via Stripe. Usage is metered for API calls above your plan's included allocation. You can upgrade, downgrade, or cancel at any time from the customer portal.

# Check your current usage via API
GET /api/usage
Authorization: Bearer YOUR_API_KEY

{
  "plan": "swarm",
  "period_start": "2026-04-01",
  "api_calls_used": 12840,
  "api_calls_included": 50000,
  "overage_cost": "$0.00"
}

Need help?

Join the MoltBot community or reach out to support.

Join Community Full Tutorial →