Kortix is a cloud-based operating system that runs AI agents to automate company operations — from coding and API integrations to document processing and infrastructure tasks. It’s designed for founders, engineers, and ops teams who want to replace manual workflows with autonomous agents that work 24/7. Built on OpenCode, it provides a full Linux sandbox with persistent memory, enabling agents to learn and improve over time.
Technically, Kortix uses TypeScript for its frontend and API, Docker for containerization, and OpenCode as its agent runtime. It supports deployment on local machines, VPSs (like Hetzner or JustAVPS), or via Kortix Cloud. The system integrates with 3,000+ tools via OAuth, REST, CLI, and MCP servers, and all agent configurations are stored as Markdown files with version-controlled triggers and skills.
What You Get
- AI Agents as Markdown Files - Each agent is defined in a Markdown file with identity, permissions, tools, and triggers — enabling human-readable, version-controlled configuration of autonomous workers.
- Persistent Memory with Semantic Search - All agent decisions, preferences, and context are stored on the filesystem and searchable semantically, allowing agents to retain and build upon past knowledge.
- 60+ Built-in Skills - Pre-built knowledge packs for coding, browser automation, legal writing, spreadsheet analysis, deep research, and more — directly usable by agents without custom development.
- Autowork Execution Loop - Agents automatically continue working until they can verify a task is complete, reducing human intervention and ensuring task fidelity.
- Cron & Webhook Triggers - Schedule agents to run at specific times or trigger them via HTTP webhooks, Git events, or other system events — all defined in agent frontmatter.
- Multi-Channel Access - Monitor and interact with agents via web dashboard, Slack, Discord, Telegram, MS Teams, iOS, and Android apps — no need to be at a computer to manage operations.
Common Use Cases
- Running a 24/7 customer support system - A SaaS company deploys a Kortix agent to monitor Gmail and Slack for support tickets, auto-research solutions using knowledge bases, and respond with templated fixes — escalating only when needed.
- Automating financial reporting - A startup’s bookkeeping agent pulls data from Stripe, QuickBooks, and Google Sheets daily, generates PDF reports, and emails the CFO with key metrics and anomalies.
- Automated engineering on-call rotation - An engineering agent monitors GitHub commits and CI/CD failures, auto-reverts broken builds, runs diagnostics, and posts summaries to Slack — reducing burnout and response time.
- Onboarding new hires autonomously - A recruiter agent triggers on new HR form submissions, creates Notion pages, invites to Slack, provisions AWS accounts, and schedules orientation calls with calendar integrations.
Under The Hood
Architecture
- Monorepo structure with clear separation of concerns, isolating applications, core services, and shared packages into modular, independently deployable units
- API layer decouples routing from business logic using type-safe Hono.js routes with Zod validation and structured responses, enhanced by service imports for domain-specific operations
- Plugin-based core system with wildcard pattern matching enables dynamic, extensible sandbox behavior without monolithic dependencies
- Dependency injection is achieved through workspace links and Dockerized service isolation, allowing independent development and scaling of database, agent, and runtime components
- Runtime and sandbox management abstracted into dedicated services with explicit state enums and result schemas, improving testability and control
- Frontend and mobile clients use consistent UI libraries with client-side hooks, maintaining design integrity while strictly decoupling from backend concerns via well-defined API contracts
Tech Stack
- Hono.js as the primary HTTP framework, paired with Drizzle ORM and PostgreSQL for type-safe data persistence
- Multi-platform development powered by Bun and TypeScript, unifying API, web, and mobile applications within a single monorepo
- AI capabilities integrated via ai-sdk with multiple LLM providers, complemented by Stripe for billing and Dockerode for container orchestration
- Mobile application built with React Native and React 19, leveraging native components and rich animations for immersive experiences
- Infrastructure managed through Docker Compose with environment-aware configurations, supporting seamless dev-to-prod transitions
- Session state and agent reasoning tracked through structured JSON logs embedding model metadata and decision traces
Code Quality
- Extensive test coverage spanning unit, integration, and end-to-end scenarios with clear separation of mocks and full-stack validations
- Consistent, domain-driven naming and structured error handling across APIs, reinforcing clarity and maintainability
- Strong type safety and modular isolation in test helpers, ensuring deterministic behavior for database and file system interactions
- Comprehensive linting and standardized test structures promote uniformity, with centralized E2E orchestration and environment-aware execution
- Security-first testing includes proactive validation of JWT integrity, token boundaries, and attack vector simulations
What Makes It Unique
- Native integration of Remotion enables dynamic, timeline-synchronized video generation directly from AI-generated code, eliminating external tooling dependencies
- Domain-specific skill modules function as executable knowledge graphs, embedding structured patterns for legal, design, and automation tasks to guide AI behavior
- Configurable sandbox environments allow simultaneous development and self-hosted instances on a single machine, solving a key pain point in local AI workflows
- AI-driven deployment pipelines auto-generate semantic subdomains and filter binaries via real-time file analysis, ensuring only source code is deployed
- Browser automation tooling captures video and debugging data tied to agent actions, creating self-documenting, verifiable AI workflows
- Automated design system enforcement via palette generation and typographic rules embedded in skill files, preserving human-crafted aesthetics in AI-generated outputs