Vellum vs ChatGPT: assistant-shaped, or an open-source AI Operating System for a team

What each product is built to be

Vellum is a personal AI assistant for one person. ChatGPT is a chat assistant on OpenAI's cloud, bought as a subscription per person. Kortix is the open-source AI Operating System: your agents, their skills, your company memory, and every connector in one git repo you own. The first two are built for a person to ask; Kortix is built to be owned and governed by a team. The Vellum alternatives comparison covers the wider field.

Whose memory is it?

Memory is where the two assistants differ least and matter most. Vellum builds long-term memory around you and keeps its credentials vault on your machine or in its cloud. ChatGPT's own pricing page says content is used to train its models with an opt-out available, and lists memory as limited, standard or expanded by plan, checked October 2026. Kortix keeps company memory as plain files in one git repo you own, so it is versioned, diffable and reviewable by the team rather than held in a vendor's account.

Who sets the rules on a tool call

A personal assistant leaves tool rules to the person using it. A team needs rules it can write down. In Kortix, connector and agent grants live in the project's kortix.yaml, and each connector call is set to run, ask or block, down to its arguments:

connectors:  - slug: gmail    provider: composio    app: gmailpolicies:  - match: "get_*"    action: always_run  - match: "*"    action: require_approval

OpenAI puts group permissions and admin controls on business plans rather than individual ones, and Vellum's fine-grained permissions are built for one person's data.

Where the work runs and who reviews it

Kortix runs every session on its own isolated cloud computer, so thousands run in parallel on one config. Session work reaches main through a change request a person approves, and merge is default-deny for agents. Vellum runs your assistant locally or in Vellum Cloud, and ChatGPT runs in OpenAI's cloud. For a team, the review step is the difference between work landing in the business and work waiting for a person to read the diff.

How to choose

Pick ChatGPT for individual chat. Pick Vellum for one person who wants a private assistant with long-term memory and a model choice. Pick Kortix when the work belongs to a team that needs one shared memory, scoped tool permissions, isolated sessions and a review step. Self-host is free, or run it on Kortix Cloud.