Kortix AI Management System v0.13.0 Released

What changed

Kortix AI has announced the latest release, v0.13.0, of its open-source AI Management System. This system is designed to manage AI agents, their associated skills, company memory, and connectors within a single, version-controlled git repository. The platform aims to provide a self-hostable and auditable alternative to commercial offerings such as Claude Cowork and ChatGPT Work. Key features include the ability to run agents on dedicated cloud computers, with each session isolated in a sandbox environment on its own branch. Work produced by agents is submitted through a change request system, requiring human approval before being merged into the main codebase. The system supports any AI model provider through API keys, allowing users to leverage their existing subscriptions or deploy their own models. Kortix offers flexibility in deployment, supporting managed cloud hosting at $40 per seat per month plus usage, or self-hosting on private infrastructure, including VPCs or on-premise networks, for free.

Core Components Managed:

• Agents: OpenCode agents with scoped tool access, designed for specific roles or tasks, capable of self-modification.
• Skills: Reusable knowledge modules that encode company-specific processes, shareable across sessions.
• Memory: A persistent company knowledge base, starting with plain files and designed to compound learning over time.
• Connectors: Integrations with over 3,000 applications, plus support for MCP, OpenAPI, and GraphQL. Credentials are managed server-side to prevent exposure within agent sandboxes.
• Secrets: Encrypted secrets are granted per agent and injected as environment variables at runtime.
• Channels: Integration with communication platforms like Slack (with Microsoft Teams and email/voice in development) allows agents to be invoked directly from team workflows.
• Triggers: Automated session spawning via cron jobs or signed webhooks.

Workflows can be executed on-demand, human-assisted, or fully automated based on schedules or triggers.

Why it matters for builders

For AI builders, Kortix provides a framework to treat AI agent development and deployment as a codebase. The open-source nature and self-hosting capabilities empower teams to maintain full ownership of their AI infrastructure, data, and models, mitigating vendor lock-in. The ability to integrate any model provider and manage agent configurations via git offers significant flexibility. Furthermore, the system's emphasis on versioning, diffing, and reviewable change requests for agent outputs promotes a more robust and governable AI development lifecycle, akin to traditional software engineering practices.

Practical impact

Developers can begin using Kortix by installing its CLI tool with a single command (curl -fsSL https://kortix.com/install | bash). A kortix init command scaffolds a new project, setting up agents, skills, and runtime configurations within a git repository. The kortix ship command deploys the project to the cloud. For immediate use without installation, users can sign up at kortix.com to create a project and start a session. The CLI allows for creating new sessions (kortix sessions new --prompt "Summarize this week's commits and open a change request"), listing change requests (kortix cr ls), and interacting with agents via terminal (kortix chat). For self-hosting, the kortix self-host start command initiates a local instance, and the CLI can be switched between cloud and self-hosted environments using kortix hosts use cloud or kortix hosts use selfhost.

Caveats and source limits

The provided source mentions a "fresh release" and lists the latest version as v0.13.0, but does not specify the exact release date for this version, only a published_at date in 2026 for the repository metadata. Pricing for the managed cloud is listed as $40/seat/mo + usage, but specific details on usage-based costs or tiered plans are not elaborated. While the system supports "any provider" for models, specific examples of supported third-party model APIs beyond general mentions of OpenAI and Anthropic are not detailed. The source also notes that competitor row information was based on publicly documented behavior as of July 2026, implying potential for changes. The exact performance benchmarks or independent reviews of the system's capabilities are not included in the provided excerpts. The license is listed as "Other" or "NOASSERTION", which may require further investigation for commercial use implications. The excerpt mentions "5 AI signals, 5 developer signals" but does not detail what these signals represent. The source also states "The whole platform ships under one version (root VERSION) — API, frontend, CLI and desktop release together as vX.Y.Z", indicating a monolithic release strategy for components.