# Copilot Studio vs Kortix: an open-source head-to-head

> Microsoft Copilot Studio builds agents inside the Microsoft stack; Kortix, the open-source AI Operating System, runs each agent from a git repo you own and lands its work through a change request. Here is how the two compare, dimension by dimension.

URL: https://opensourcecopilotalternative.com/copilot-studio-vs-kortix

## An agent is a file, and what it may reach is a block in the same repo

`kortix.yaml`

```yaml
# kortix.yaml
agents:
  kortix:
    file: agents/kortix.md
    connectors: all
    secrets: all
    skills: all
    kortix_permissions: all
```

Source: https://kortix.com/docs/project/manifest

The definition and its grants sit in one `kortix.yaml` a person can read, diff and review. Kortix is open source, so the file is yours.

## Where Copilot Studio and Kortix each put the agent layer

One column states what Microsoft documents; the other is Kortix, the open-source option we recommend.

| | Kortix (recommended) | Copilot Studio |
| --- | --- | --- |
| Open source | Yes | No |
| Where the agent's definition lives | Files in a git repo you own | In the studio, on Microsoft's platform |
| Where it runs | Kortix Cloud, your VPC, or on-prem | Microsoft's cloud (Power Platform) |
| How a change is reviewed | A change request a person approves; merge is default-deny for agents | You publish from the studio; admins govern from Power Platform |
| Model choice | Any model provider with your own keys | Three harnesses; Microsoft Foundry and Azure AI Search models |
| Reach beyond Microsoft tools | 3,000+ apps in a click, plus MCP, OpenAPI, GraphQL, raw HTTP | Power Platform connectors, REST APIs, MCP servers |
| Self-host | Yes, self-host is free | No |
| How you buy it | A free tier, then $40 per seat per month | A $200 per month capacity pack, or pay-as-you-go |

Kortix cells come from kortix.com/docs; Copilot Studio cells come from Microsoft Learn's Copilot Studio overview, agents, publish and billing pages, checked October 2026.

## What Microsoft Copilot Studio is

Microsoft Copilot Studio is a graphical, low-code studio for building and managing AI agents and workflows on Microsoft's platform. A builder describes an agent in plain language, connects it to organizational data, tests it, and publishes it to channels such as Microsoft Teams, a website, or Microsoft Copilot. Microsoft documents three harnesses for these agents: a GitHub Copilot harness for reasoning-heavy, multistep work, a standard harness for rule-based conversations, and a Copilot chat harness that extends Microsoft Copilot Chat [Microsoft Copilot Studio overview](https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio).

Agents can call Power Platform connectors, REST APIs, MCP servers, and other agents, and they run on Microsoft's cloud inside a Power Platform environment. The strength is real: a team that already lives in Microsoft 365 gets a fast path to agents inside the apps and identity it already runs. The trade is where the definition lives: the agent, its instructions and its reach sit in the studio on Microsoft's platform, not in a repository the team controls.

## What a team owns when the agent layer is a repo

Kortix is the open-source AI Operating System, and it puts the agent layer in a git repo the team owns. An agent is `agents/<name>.md` (behavior) plus an `agents.<name>` block in `kortix.yaml` (governance). One isolated sandbox per session: each session has its own isolated machine and branch. Thousands of agents run in parallel on one config, each on its own cloud computer [the Kortix docs](https://kortix.com/docs).

The review model follows from that. Session work reaches `main` through a change request. Merge is default-deny for agents, so a person reviews the diff before it lands. You can scope an agent's reach in the same file: set each tool to Allow, Ask or Block, down to the arguments of each call. Approval gates are off until you set them. Reach is not limited to one vendor's stack: Kortix connects 3,000+ apps in a click, plus MCP, OpenAPI, Postman, GraphQL and raw HTTP. Connector credentials are brokered server-side and never enter the machine.

A Kortix agent's model choice is open too: any model provider with your own keys; the ChatGPT plan you already pay for; or sign in with your OpenCode Console account for OpenCode Zen and Go. You also place it where it must run: run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free. If your security review needs the agent layer on your own network, that is the lane Kortix is built for.

## When to pick which

Pick Copilot Studio when the agents must live inside Microsoft's tools and Microsoft's cloud, and the builders are already makers in the Power Platform. It is the shorter path to an agent that answers inside Teams or SharePoint, and its administration runs through tools an IT team already knows.

Pick Kortix when the agent layer itself is the thing you want to own. That fits a team that needs the configuration in a git repo it can read and diff, each change reviewed through a change request, agents reaching tools outside Microsoft, and a deployment on its own network or VPC. For the switching intent behind this page, see the wider [open-source Copilot alternatives](/best-open-source-copilot-alternatives), the direct [Kortix vs Copilot](/kortix-vs-copilot) breakdown, and how agent governance works in a repo on [managing Copilot agents](/copilot-agent-management). The same argument, in long form, is in the [Kortix](https://kortix.com/blog/open-source-microsoft-copilot-alternative) guide.

_FAQ_

## Copilot Studio questions, answered

### Can I move agents off Copilot Studio?

A Copilot Studio agent does not export as a portable file, because its definition lives in the studio. A team rebuilds the logic as Kortix agents, where each agent is an `agents/<name>.md` file plus a governance block in `kortix.yaml`. The prompts and the job carry over; the studio configuration does not, so plan a rebuild rather than a migration.

### Does Copilot Studio lock me to Microsoft tools?

Copilot Studio reaches outside Microsoft through Power Platform connectors, REST APIs and MCP servers, so it is not closed to other tools. What stays on Microsoft's side is the agent definition and the runtime: the agents run in Microsoft's cloud on a Power Platform environment, and the configuration lives in the studio, not in a repo your team controls.

### Can the alternative be self-hosted?

Yes. Kortix is open source, and self-host is free. Run it on Kortix Cloud, in your VPC, or on your own on-prem network. The stack starts with `kortix self-host start`, one Docker Compose stack on a server you control; the first run asks six things and no model key, and models are added in the app afterward.

### Where does the agent's configuration live in each?

In Copilot Studio it lives in the studio, on Microsoft's platform, and admins govern agents from the Power Platform admin center. In Kortix it lives in a git repo you own: `agents/<name>.md` for behavior and an `agents.<name>` block in `kortix.yaml` for governance, reviewed like any other change.

## Run the agent layer from a repo you own

Start with one job and one agent.

[Try Kortix](https://kortix.com) · [See the Copilot Studio alternatives](https://opensourcecopilotalternative.com/copilot-studio-alternatives)