# Maritime Hosting platform for persistent AI agents with isolated VMs, persistent storage and automatic sleep and wake. Canonical profile: https://agenten-kompass.mmeierhoff.chatgpt.site/en/agents/maritime/ Editor: [Michael Meierhoff](https://agenten-kompass.mmeierhoff.chatgpt.site/en/about/) Sources reviewed: 2026-10-10 Basis: provider documentation. Examples are editorial suggestions, not hands-on product tests. ## Overview Maritime provides the runtime environment; the selected agent framework determines capabilities and tool access. - Provider: Maritime AI, Inc. - Product family: Maritime - Profile type: Hosting platform for AI agents - Environment: Managed cloud; dashboard, CLI, REST API and SDKs - Connections: Agent templates, Docker containers, Telegram, and TypeScript and Python SDKs - Setup: Create an account, choose an agent template or your own container and configure model access. - Pricing model: Free: 3 agents; Starter: USD 20/month. Model usage billed separately. ## Operation & oversight Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured. Deployment covered: Managed Maritime runtime; agent behavior depends on its framework Sources reviewed: 2026-10-10 ### Files and work state Documented with conditions: The persistent volume, usually /data, survives restarts and deployments; other paths are disposable. Evidence: https://maritime.sh/docs/how-it-works ### Authentication Documented with conditions: Environment variables are stored encrypted. The validity of specific browser logins remains open in this source. Evidence: https://maritime.sh/docs/how-it-works ### Starting further runs Documented with conditions: Messages, platform triggers, web requests and file/exec calls wake agents. In-process timers pause during sleep. Evidence: https://maritime.sh/docs/how-it-works ### Human intervention Open in reviewed sources: A uniform action approval mechanism across agent frameworks is not established here. Evidence: https://maritime.sh/docs/how-it-works ### Detecting errors Documented: The dashboard exposes logs. Evidence: https://maritime.sh/docs/how-it-works ### Recovering from errors Open in reviewed sources: Sleep/wake preserves state. Recovery from arbitrary application errors remains open. Evidence: https://maritime.sh/docs/how-it-works ### Editorial assessment The managed runtime reduces infrastructure setup. Storage paths, schedules and framework error behavior remain configuration tasks. ## Provider Provider of the managed Maritime platform for running agents. - Operator: Maritime AI, Inc. - Organization: Company - Website: https://maritime.sh/ - Reviewed: 2026-10-10 - Provider source: https://maritime.sh/ - Provider source: https://maritime.sh/pricing ## Prices and included usage Pricing checked: 2026-10-10 Free: 3 agents; Starter: USD 20/month. Model usage billed separately. ### Free USD 0 3 agents, with 1 awake at a time. Price source: https://maritime.sh/pricing ### Starter USD 20/month 20 machines, with 5 awake at a time. Price source: https://maritime.sh/pricing ### Growth USD 100/month 100 machines, with 25 awake at a time. Price source: https://maritime.sh/pricing Hosting and model usage are billed separately. Larger plans and paid resource upgrades are available. Always-on costs an additional USD 20 per machine per month on the listed paid plans. Pricing details: https://maritime.sh/pricing ## Key features ### Isolated runtime Each agent runs in its own VM. Source: https://maritime.sh/docs/how-it-works ### Persistent storage The persistent volume is usually mounted at /data and survives restarts and deployments. Source: https://maritime.sh/docs/how-it-works ### Events wake agents Messages, platform triggers, web requests and file or exec API calls wake sleeping agents. Source: https://maritime.sh/docs/how-it-works ### Manage agents programmatically Dashboard, CLI, REST API and SDKs support provisioning and management. Source: https://maritime.sh/docs/how-it-works ### Templates and custom containers An existing agent framework or a custom Docker container supplies the agent logic. Source: https://maritime.sh/docs/how-it-works ## Requirements and limits ### Choose storage paths deliberately Limitation: Files outside the persistent volume are disposable. Source: https://maritime.sh/docs/how-it-works ### Configure schedules on the platform Limitation: In-process timers do not run during sleep. Recurring tasks need platform triggers. Source: https://maritime.sh/docs/how-it-works ### Account for model access Requirement: Model calls use separate AI credits or your own provider keys. Claude Code requires an Anthropic key. Source: https://maritime.sh/docs/how-it-works ### Test login persistence Not yet confirmed: Persistence of specific browser logins and recovery after failures have not yet been tested by the editor. Source: https://maritime.sh/docs/how-it-works ## Example workflows ### Run recurring research - Your input: An agent framework, sources and a platform schedule. - Possible result: An agent gathers results and saves them to the persistent volume. - What to check: Check sources, storage paths, recovery and model costs. ### Provision an agent per user - Your input: Your own application connected to the Maritime SDK. - Possible result: Separate agent instances for application users. - What to check: Check permissions, data isolation and concurrently active instances. ## Context Cloud hosting; persistence depends on the storage path. Plans limit concurrently active machines. Official product documentation reviewed. Workflows and assessments are editorial; hands-on product tests are outside the scope of this directory. ## Sources - https://maritime.sh/ - https://maritime.sh/docs/how-it-works - https://maritime.sh/pricing [Sources and method](https://agenten-kompass.mmeierhoff.chatgpt.site/en/sources/)