ChatGPT Work vs an open-source alternative your company can own

ChatGPT Work is OpenAI's agentic mode for getting finished work out of ChatGPT; Kortix is the open-source alternative a company can run on its own infrastructure. Kortix is the open-source AI Management System: one git repo holds your agents, skills, memory and connectors, with any model and your own keys, self-hosted or run on managed cloud. This is what ChatGPT Work does well, where it stops when a whole company has to own the result, and what an open-source platform gives you instead.

What ChatGPT Work gets right

ChatGPT Work is a capable agent. OpenAI launched it on July 9, 2026 as a mode inside ChatGPT, powered by GPT-5.6, and describes it as an agent that takes action across your apps and files, stays with a project for hours, and turns a goal into finished work (OpenAI announcement).

• It finishes the work. Give it a goal and it returns a document, spreadsheet or presentation.
• It runs for a long time. A task can stay with a project for hours.
• It lives where your team already is. ChatGPT Work is a mode inside ChatGPT, with the same login and billing.
• The model and the agent ship together. GPT-5.6 is built by OpenAI beside the agent, so the two are tuned as one.

Where it stops for a company

ChatGPT Work is closed source and runs in OpenAI's cloud, with no self-host (Kortix on GitHub). It runs OpenAI's models, its configuration lives inside OpenAI's product, and access comes through paid, usage-metered plans. A company that needs to own its own platform has to look elsewhere.

• You cannot read or change the code. The platform is closed, so you take the behavior OpenAI ships.
• You cannot self-host. It runs in OpenAI's cloud, not your VPC or your on-prem network.
• You get one vendor's models. GPT-5.6 is the model, so model choice and model cost are OpenAI's to set.
• Configuration stays in their product. Agents, instructions and context sit in OpenAI's system, where there is no repo to diff or roll back.
• Access is metered. You pay for usage on OpenAI's plans.

What an open-source platform gives you

Kortix is the open-source platform a company can own end to end. Everything the company runs lives in one git repository: agents, skills, memory, connector configuration and triggers are text files with an author, a timestamp and a diff.

• One repo is the company. Agents, skills, memory and triggers are files you can grep, diff and roll back with git.
• Any model, your keys. Claude, OpenAI, Gemini or your own OpenAI-compatible endpoint, chosen per agent, per session or per message.
• Self-host where you want. Run Kortix on a laptop, a VPS, your VPC or on-prem, or on the managed cloud.
• Every tool the company uses. 3,000+ apps plus MCP, OpenAPI, GraphQL and raw HTTP, with credentials brokered server-side so raw keys never enter the machine.
• A real agent harness. Kortix runs the OpenCode harness, so agents plan, use tools and finish multi-step work, with allow, ask or block on every tool call.
• Every session gets its own computer. An isolated Linux machine boots per session, thousands run in parallel, and nothing runs on your laptop.
• One gate to land work. Wherever a session starts, the result lands as a change request a person reads as a diff before it reaches main.

You can also sign in with the ChatGPT subscription you already pay for, so moving to an open-source platform does not mean paying for models twice.

Configuration that lives in the repo

A Kortix project is a git repository, so the platform's behavior is text you can review before it runs. A single manifest declares the machine image, the connectors, the triggers and what each agent may touch; everything past it is markdown.

# kortix.yaml: the company, as files
agents:
  support-triage:
    file: agents/support-triage.md
    connectors: all
triggers:
  - slug: daily-digest
    type: cron
    cron: "0 0 9 * * 1-5"

An omitted grant resolves to none, so least privilege is the default, and a change to any of it is a diff a person approves. ChatGPT Work keeps the same settings inside OpenAI's product, where you can read them but not version them.

Side by side

When to pick which

ChatGPT Work suits a person or a small team that is happy inside ChatGPT, on GPT-5.6, and does not need to own or self-host the platform.

Most teams keep ChatGPT Work for fast thinking and move the repeatable work to Kortix: the jobs that run on a schedule, reach your own tools, and land through review. Those jobs need a platform the company can version, audit and roll back, which is what the repo gives you.

Buyer's decision guide

Five questions decide the choice: where the platform runs, which models you can use, who holds the configuration, how finished work reaches the rest of the company, and what the cost looks like as the team grows.

• Self-hosting. ChatGPT Work has no self-host. Kortix runs on your laptop, a VPS, your VPC or on-prem, or on the managed cloud.
• Model choice. ChatGPT Work runs GPT-5.6 only. Kortix runs any model, with your own keys, per agent or per message.
• Ownership of configuration. In ChatGPT Work the agents and context live in OpenAI's product. In Kortix they are files in a git repo you own.
• How work lands. ChatGPT Work returns a deliverable to ChatGPT. Kortix opens a change request you review as a diff before it reaches main.
• Cost shape. Kortix is free to start and free to self-host; the managed cloud adds per-seat pricing with usage on top.

If you are also weighing the desktop assistants, Kortix vs Claude Cowork and what Kortix actually is go deeper. For the wider category, read the open-source Microsoft Copilot alternative.

Start with one job on a platform your company owns. Get started with open-source Kortix or read the docs.