- Open source
- Yes
- Hosting
- Kortix Cloud, your VPC, or on-prem
- Where the model comes from
- Any model provider with your own keys
- Where the code lives
- The project repo, one branch per session
- Where agents and memory live
- One git repo you own
- Approvals
- Review cards in the thread, join policy per channel
- Cost
- Self-host free; Kortix Cloud runs on credits
- Channels
- Slack and Microsoft Teams are live
Alternatives to Claude Tag, compared on ownership
The open-source Claude Tag alternative, judged by what you own
Claude Tag gives a team a capable teammate, and Anthropic keeps the runtime, the memory and the model. Kortix is the open-source AI Operating System: your agents, their skills, your company memory, and every connector in one git repo you own, with the agents working on real cloud computers.
The open-source AI Operating System
channels
One Slack thread is one session.kortix channels connectSend /kortix login in Slack or /login in TeamsThe agent replies in the Slack thread or Teams chat, with live progress.A DM session is private to you.A decision on a review card or a connector approval card resumes the agent under every policy.
Run it on Kortix Cloud, in your VPC, or on your own on-prem network. Self-host is free.
Open source · Any model provider with your own keys · Slack and Microsoft Teams are live
What each option leaves in your hands
The difference that decides the purchase is who owns the machine, the repo and the model.
- Open source
- No
- Hosting
- Anthropic's platform, inside Slack
- Where the model comes from
- Claude models only
- Where the code lives
- Codebases an admin connects
- Where agents and memory live
- Per-channel memory, owners can edit
- Approvals
- Admins set tool access and spend caps
- Cost
- Usage-based, on top of Team or Enterprise
- Channels
- Slack only today
- Open source
- Yes
- Hosting
- A server and machines you operate
- Where the model comes from
- Whichever CLI runtime you configure
- Where the code lives
- Folders you bind on your machines
- Where agents and memory live
- Files on each agent's state directory
- Approvals
- Roles and capability checks per action
- Cost
- Free to run yourself; you supply the runtime accounts
- Channels
- Its own web workspace
Three jobs this choice actually decides
Each card is the work that comes back to the person who asked for it.
Slack thread to change
A thread that ends in a diff someone ships
A mention starts a session on its own cloud computer, and the change request posts back into the same thread with Ship it, Reject and Ask for changes.
session-branch.diffWhat comes back
export async function retry(fn, n = 3) { export async function retry(fn, attempts = 3) { if (!Number.isInteger(attempts) || attempts < 1) throw new RangeError("attempts must be positive");
The session opened the change request; a person holding project.gitops.merge ships it.
Bounded tools
Allow, Ask or Block, call by call
Each connector action carries a policy, and an Ask call posts a card with the arguments before anything runs.
gmail-policy.checksWhat comes back
- gmail sendAsk
- github mergeBlock
- notion readAllow
- join policyowner_approval
An Ask call holds the agent's turn, and the decision resumes it.
One repo, many doors
The same repo in Slack, Teams, the web and the CLI
One Slack workspace or Teams tenant can serve several projects, and each channel binds to one project, so the teammate in Teams and the engineer in the terminal work from the same repo.
surfaces.sheetWhat comes back
Slack
- Starts a session
- A mention in a thread
- Work lands in
- The same thread
Teams
- Starts a session
- A post in a channel
- Work lands in
- The same chat
Web and CLI
- Starts a session
- A session you start
- Work lands in
- The project's change requests
Join policy decides who may enter a session: project_open, owner_approval or owner_only.
Open source. Read it, fork it, audit it.
The permission model, the connector policies and the review gate are code you can open and check yourself.
The three questions people ask first
What are some alternatives to Claude Tag?
Kortix and open-tag. Kortix is the open-source AI Operating System, with agents, skills and memory in a git repo the company owns. open-tag is a self-hosted workspace for coding agents. Both run on infrastructure the team controls, which is the part Claude Tag keeps.
Does Claude Tag cost extra?
Yes. Claude Tag is consumption-based and billed to your organization on top of a Team or Enterprise plan, with an organization-wide cap and per-channel limits that admins set. Direct messages bill to the person's own Claude account instead.
How useful is the Claude tag?
Very, if your team already works in Slack on a Team or Enterprise plan. Anthropic says 65% of its product team's code comes from its internal version, and it remembers channel context, schedules follow-ups and works asynchronously. It is Slack only today, on Claude models, in beta.
Own the machine the agent runs on.
Connect Slack or Teams to one project, then start with a single thread.