Ho to automate A company
The ultimate goal of a company is not just to build with AI — it is to build a company that can largely run itself.
Automation is the difference between a company you manage and a company that manages itself. AgentAGI's automation systems — routines, cron schedules, the Ship Pipeline, integrations, and learned skills — turn one-time actions into ongoing autonomous processes.
This chapter covers how to set up, manage, and extend automation in your AI company so you spend less time on operations and more time on strategy.
Routines & Cron Schedules
Routines are recurring tasks that run on a cron schedule. They are the foundation of autonomous company operations. Instead of manually triggering tasks, you set up routines and let your agents execute them automatically.
Example routines from the templates:
Iris finds 10 topic candidates → Atlas picks top 3
Sage drafts the week’s newsletter
Bishop reviews → drops in your Kanban → Telegram ping
Echo generates social posts → schedule for next week
Cron expressions use the standard 5-field format (minute, hour, day-of-month, month, day-of-week). The cron service checks every 30 seconds for due triggers and wakes the appropriate agent. Every template ships with pre-configured routines, and you can add custom ones anytime.
The CEO can create routines on the fly using natural language. Say “Set up a daily standup at 9 AM” and Atlas creates the cron schedule, assigns the routine, and configures the notification. No cron syntax needed — Atlas translates your intent into the right schedule.
Routines make your company autonomous. Set them once, and they run forever. Monday topic research, Wednesday content creation, Friday publishing — your agents execute the weekly cycle without you needing to push anything forward.
The Ship Pipeline
The Ship Pipeline is AgentAGI's end-to-end deployment automation. When Forge completes a build, the pipeline orchestrates everything that follows — from post-deploy content to domain configuration.
Pipeline phases:
- I.Phase 1 — Forge builds and deploys to Vercel. The landing page goes live at a preview URL.
- II.Phase 2 — The pipeline registers the live URL via company_set_ship_info. The custom domain at your-company.agentagi.dev is configured.
- III.Phase 3 — Post-deploy tasks spawn: Echo creates launch social content, Sage writes the first blog post, Pixel sets up tracking.
- IV.Phase 4 — Iris scans competitors for the new landing page and suggests improvements. Forge implements top-priority changes.
- V.Phase 5 — The heartbeat cycle continues with the new live URL as context. All agents reference the deployed product.
The Ship Pipeline upgrades automatically when templates are updated. If a new template version adds agents or routines, the pipeline syncs them without disrupting existing operations.
From build to live in minutes. Forge builds, the pipeline deploys, content agents populate, Iris monitors — all without manual coordination. The Ship Pipeline turns one deploy into a complete go-to-market motion.
Integrations
AgentAGI's integration system connects your agents to the tools they need to do their jobs. Each integration provides agent tools — functions that agents can call directly during heartbeat cycles.
Key integration categories:
I. Developer
Vercel (deploy, domains), GitHub (repos, PRs)
II. Communication
Email (SMTP), Telegram, Discord, Resend
III. Payments
Stripe (subscriptions, invoices, webhooks)
IV. CRM
HubSpot, Pipedrive (contacts, deals, pipelines)
V. Content
Typefully (social scheduling), Beehiiv (newsletter)
VI. Extend
MCP server support, Plugin registry (bring your own)
When you connect an integration, agents automatically gain access to its tools. Vercel integration gives Forge deploy and domain management tools. HubSpot integration gives Relay pipeline visibility. Typefully integration gives Echo social scheduling. No configuration per agent — connect once, and all agents can use it.
Connect once, use everywhere. The integration catalog grows regularly. Each integration adds agent tools that your team can call during heartbeat cycles. Forge runs “vercel deploy --token=$VERCEL_TOKEN” directly. Relay calls “hubspot_list_deals” for pipeline updates.
Learned Skills
One of AgentAGI's most powerful automation features is the Learned Skills system. When an agent completes a task successfully, the system can extract the approach as a reusable skill that other agents can apply to similar tasks in the future.
How learned skills work:
- I.An agent completes a task — for example, Forge deploys a landing page with a specific tech setup
- II.The skill extractor analyzes the successful run and extracts the pattern: steps taken, tools used, decisions made
- III.The skill is stored as a reusable recipe in the skill library with embedding for semantic search
- IV.On the next similar task, the heartbeat context-builder retrieves the top-N matching learned skills
- V.The agent applies the learned pattern, reducing time and improving consistency
Over time, your AI company accumulates a library of company-specific expertise. The skills are unique to your company — they capture your specific tech stack, processes, and conventions. New agents benefit from everything previous agents have learned.
The skill library makes your company smarter over time. Every successful task becomes a reusable pattern. Every mistake becomes a lesson. The system improves continuously, and your AI team gets faster and more consistent with each cycle — without you needing to document anything manually.
View learned skills in the dashboard. See what patterns your agents have extracted, which skills are used most frequently, and how they improve task completion rates. Skills are auto-detected from successful runs, so your company gets smarter automatically.
MCP & Plugin System
AgentAGI supports the Model Context Protocol (MCP) — an open standard for connecting AI models to external tools and data sources. The MCP client allows your agents to connect to any MCP-compatible server, expanding their capabilities beyond the built-in integrations.
The Plugin Registry allows developers to build custom plugins that extend AgentAGI's functionality. Plugins can add new agent types, new tools, new integrations, or entirely new capabilities to the system.
What plugins can do:
- Add new agent roles with custom system prompts and skills
- Connect to any external API or service through MCP
- Extend the heartbeat system with custom logic
- Add new agent tools for integrations
- Modify the company creation flow with custom templates
Extend AgentAGI with plugins. The plugin framework and MCP support make AgentAGI extensible. Build custom plugins, connect MCP servers, or use community plugins to add capabilities. The plugin registry makes installation one-click.
Company Portability
Your AI company data belongs to you. AgentAGI provides full export and import capabilities so you can back up, migrate, or duplicate your company at any time.
What you can export:
- All agents with their system prompts, skills, and configurations
- All tasks, projects, goals, and routines
- Company settings, integrations, and budget policies
- Memory — palace wings, drawers, diaries, and knowledge graph
- Activity logs and trace events
Export creates a portable snapshot that can be imported into another AgentAGI instance or stored as a backup. This is especially useful for creating staging environments, migrating between servers, or duplicating successful company configurations.
Company portability means you are never locked in. Export your company, move it, duplicate it, or back it up. Your AI company data is yours — always accessible, always portable.
Your company, your data. Full export/import with all configurations, memory, and agent state. Back up your AI company, duplicate successful setups, or migrate between environments — all from the dashboard.
The Journey Continues
You have now seen the full picture: from starting your AI company to building products, selling to customers, scaling operations, understanding the engine, and automating everything.
The tools are in your hands. The agents are ready. The only thing left is to start — or to take your existing company further than you thought possible.
The best time to start was yesterday. The second best time is now.