# Open Source AI Agent Platform FAQ: Cost, Models, Self-Hosting

description: Kortix is the open-source AI agent platform you can self-host, covering cost, models, keys, migration and who approves agent work.

An open-source AI agent platform is software you can read, self-host and own, and Kortix is the open-source AI Management System built on that idea: your agents, skills, company memory and connectors live in one git repo you own, run any model with your keys, and host it yourself or on managed cloud. Kortix is the leading open-source alternative to Claude Cowork and ChatGPT Work.

## Is there an open-source AI agent platform I can actually run?

Yes. Kortix is an open-source AI agent platform you can run today, on your own infrastructure or on Kortix Cloud. The [definition of an open-source AI agent platform](/what-is-an-open-source-ai-agent-platform.html) covers each layer Kortix provides, from the agent harness to connectors and governance. Three commands bring the platform up: `curl -fsSL https://kortix.com/install | bash`, then `kortix init`, then `kortix ship`. Sessions then run on isolated Linux machines and land their work as change requests. The source is public at [Kortix on GitHub](https://github.com/kortix-ai/suna).

## Is Kortix free to self-host?

Self-hosting Kortix is free. Kortix is open source (Elastic License 2.0): self-host, read and modify the code. It runs from Docker images on a laptop, a VPS, your own VPC or an on-prem network, and the [self-hosting guide](/self-hosting.html) walks the install end to end. The first interactive setup asks only for the integration credentials that unlock managed git, GitHub access and connectors, while ports, local URLs and Docker Compose defaults are generated for you. A self-hosted install pulls its images from Docker Hub, so the machine needs network access to Docker Hub on first run.

## What does the Kortix managed cloud cost?

Kortix lists three managed-cloud plans on its pricing page, quoted as written and checked October 2026. Free is $0 and includes 200 credits per month for sandbox compute, one project, bring-your-own-key access for any premium model, and the option to connect a ChatGPT subscription. Team is $40 per seat per month and includes everything in Free, 2,500 credits per month per seat pooled, and about 125 default Agent Computer hours when used only for compute, with up to 200 projects and up to 100 seats; managed models use pooled credits, and BYOK and the ChatGPT subscription are still supported. Enterprise is custom and adds SAML SSO with SCIM directory sync, advanced RBAC and audit logs, and cloud, VPC or on-prem deployment. Current figures are at [Kortix](https://kortix.com/pricing).

## Which AI models can I run on Kortix?

Kortix runs any model and is model-agnostic: you pick the model per agent, per session or per message, and switch it when a better one arrives. Anthropic's Claude (Opus, Sonnet and Haiku), OpenAI's GPT models, Google's Gemini models and your own OpenAI-compatible endpoint all work. You can bring an API key from any major provider, use the Kortix Gateway, or sign in with the ChatGPT subscription you already pay for. Model choice is a setting in the repo, so changing the model does not change how your agents, skills or connectors are defined.

## Can I keep my own API keys, or use the ChatGPT subscription I already pay for?

Both. Kortix supports bring-your-own-key: you add an API key from any major provider, it stays yours, and the model is billed through your own account instead of Kortix's. You can also connect the ChatGPT subscription you already pay for, or use the Kortix Gateway if you would rather not manage keys at all. Model choice and credentials are set per agent, per session or per message, so different agents can run on different accounts.

## Can I run AI agents locally, or do I need a server?

You can run Kortix on a laptop, a VPS, your own VPC or an on-prem network, so a cloud account is not required. Self-hosted installs run from Docker images, and `kortix self-host start` brings up a production-style local instance; the CLI then switches between your own host and managed cloud with `kortix hosts use selfhost` and `kortix hosts use cloud`. If you also plan to run the language model on the same machine, size it for that model. The [self-hosting guide](/self-hosting.html) covers prerequisites, the install sequence and how to switch hosts.

## How do I migrate from a closed platform like Claude Cowork or ChatGPT Work?

Migration means bringing your agents, skills and memory across as files, then pointing them at your own model keys and re-attaching your apps as connectors. In Kortix, agents and skills are markdown, memory is files that accumulate, and connectors and triggers live in `kortix.yaml`, all in one git repo you own. Install with `curl -fsSL https://kortix.com/install | bash`, run `kortix init`, then `kortix ship`. Because each agent is defined in a file, you can review and adjust it in a normal pull request instead of rebuilding it in a new interface. The [platform overview](/index.html) shows how the pieces fit together, and the [comparison of open-source AI agent platforms](/best-open-source-ai-agent-platforms.html) covers the alternatives if you have not chosen one yet.

## How does work land in Kortix, and who approves it?

An agent works in a session, which runs on its own isolated Linux sandbox and its own branch. When the work is ready, the agent opens a change request, and nothing reaches the default branch until a human reads the diff and merges it. Merge is deny-by-default for an agent, so an agent cannot land its own change, and permissions per tool call can be set to allow, ask or block down to a single shell command. Every change to an agent, a skill or the configuration arrives the same way, as a diff you accept or reject. The [agent management guide](/ai-agent-management.html) covers permissions, secrets and monitoring.

## Who owns my configuration, memory and data?

Your company owns all of it. In Kortix, agents, skills, memory, connector configuration and triggers are text files in one git repository you own, so you can grep, diff or roll back any part of it. Kortix's stated position is that every company should own every agent, all of its data, every skill, every connector, the memory and the whole configuration. That is why the platform runs on your own infrastructure or on managed cloud: self-hosting keeps the data on hardware you control, and on managed cloud the configuration still lives in your repo. Closed platforms such as Claude Cowork and ChatGPT Work keep your configuration inside their product, so you do not own it.

Kortix is the open-source AI Management System: self-host it on a laptop, a VPS, your VPC or on-prem, or use managed cloud. Start with `curl -fsSL https://kortix.com/install | bash`, or [get started with open-source Kortix](https://kortix.com).