Octomind Cloud and Hub

Product Introduction

1. Definition: Octomind Cloud and Hub is a cloud-based AI agent runtime and orchestration platform. Technically, it is a managed container service pre-configured with the open-source Octomind agent runtime, providing a serverless environment for executing persistent, stateful AI agent sessions.
2. Core Value Proposition: It exists to enable developers and technical users to run long-lived, computationally intensive AI agents without managing infrastructure. Its primary value is persistent AI agent sessions that survive device disconnects, coupled with per-second billing and a built-in model gateway that eliminates API key management.

Main Features

1. Cloud Machines (Containers): Users provision isolated Docker containers ("Machines") with pre-installed Octomind. They select from predefined sizes (e.g., Small: 1 vCPU/2 GB RAM) and base images (Base, Node, Python, Rust). The container lifecycle is fully managed, with automatic suspension after configurable idle periods (5 min on Free, 15 min on Pro). Billing is granular: full compute while active, a reduced memory-only rate while idle but awake, and only storage costs when suspended.
2. Durable, Multi-Device Sessions: Agent interaction history, tool calls, and execution state are persisted in a durable session layer. This allows a user to start a session on a laptop, close it, and resume the exact same interactive session from another device like a phone. Sessions are fully replayable and never lost due to disconnection or machine suspension.
3. Integrated Model Gateway & Hub: The platform includes a built-in gateway to 21 AI models. Users authenticate once via the platform; no individual model API keys (OpenAI, Anthropic, etc.) are required for the included models. The accompanying Octomind Hub provides transparent per-token pricing and access to these models, which can also be used locally via the CLI with the same account.
4. Full Developer API & Web Terminal: Every UI action is backed by a RESTful API (OpenAPI spec available at /api/v1/openapi), enabling programmatic control for CI/CD, cron jobs, or product integration using developer keys (om_...). Alongside the chat UI, users can open a real web-based bash shell into their container for direct command-line access and custom tool installation.
5. Shared Persistent Directory: Each account has a centralized, persistent storage volume that is automatically mounted into every Machine. This holds code indexes, agent memory, and session history, enabling "index once, use everywhere" workflows and ensuring context persists across different compute containers.

Problems Solved

1. Pain Point: Agent Session Fragility. Traditional local AI agent processes terminate when the laptop sleeps or the SSH connection drops, destroying in-progress work. Octomind Cloud solves this with cloud-persistent sessions.
2. Pain Point: Infrastructure and Model Setup Overhead. Configuring a secure cloud VM, installing dependencies, managing Docker, and sourcing/rotating API keys for multiple AI models is a significant time sink. Octomind abstracts this with one-click Machines and a unified model gateway.
3. Target Audience: Software Developers & Engineers working on code refactoring, debugging, or batch analysis; Researchers & Data Scientists needing long-running analysis agents; DevOps Engineers seeking to automate tasks with AI in CI/CD pipelines.
4. Use Cases: Asynchronous Code Refactoring: Start a major codebase refactor, close your laptop, and review the completed diff and changes later from any device. Heavy Computational Tasks: Offload large codebase indexing or test suite execution to a cloud Machine to free up local resources. Persistent Research Agents: Deploy an agent to continuously monitor, summarize, or research a topic over hours or days, accessible intermittently.

Unique Advantages

1. Differentiation vs. Raw Cloud Providers (AWS, GCP): Unlike launching a generic VM, Octomind provides a pre-configured, optimized agent environment with session persistence and built-in AI model access, abstracting away sysadmin work. Differentiation vs. Other AI Agent Platforms: Focus on persistent, stateful sessions and a hybrid cloud/local model via the Hub and CLI, rather than stateless, single-turn task completion.
2. Key Innovation: The separation of compute (Machine) from state (Session & Shared Directory). This allows cheap, ephemeral containers to be spun up/down or suspended while preserving the agent's memory and context, enabling true per-second billing for long-running tasks. The open-core model, with the Apache 2.0 licensed runtime, provides transparency and self-hosting options.

Frequently Asked Questions (FAQ)

1. How much does Octomind Cloud cost when I'm not using it? Costs drop significantly when idle. An active Machine bills per-second for compute. When idle but awake, it bills a reduced memory-only rate (~$0.01/hr for Small). After the idle timeout (5-15 min), it suspends, incurring only storage costs ($0.10/GB-month beyond the 10GB free tier). Live spending caps (4-hour, weekly, monthly) prevent surprise bills.
2. Can I use my own AI models or API keys with Octomind Cloud? The platform's built-in gateway provides curated models without user-managed keys. For premium models like Claude or GPT-4, usage draws from prepaid credits within the platform. To use a personal API key, you would run the open-source Octomind runtime locally or on your own infrastructure via the CLI and Hub.
3. What happens to my data and work if I cancel my subscription? Upon cancellation, your account and data are not immediately deleted. A 30-day grace period is provided where all data (Machines, Sessions, Files) is retained. If you resubscribe within this period, everything is restored. You can also download all your files from the Files tab at any time.
4. Is Octomind Cloud suitable for running production workloads or CI/CD jobs? Yes, via its comprehensive Developer API. You can programmatically create a Machine, execute an agent task via a session, stream results, and terminate the Machine. This is designed for automation use cases like CI, cron jobs, or integrating AI agent functionality into your own product, using dedicated developer keys.
5. How does the performance of Octomind's included models compare to OpenAI or Anthropic? According to the platform's benchmark on real-PR tasks, its configured open models solved 24/25 tasks, claiming superior performance to Claude Code and Codex. The Pro plan provides access to these "strong open coding models" with generous caps. For top-tier performance, premium models (Claude, GPT) are available via prepaid credits.