What we believe
Kody is the shared cloud behind your team’s agents. These are the beliefs that shape it, and where each one shows up in the product today.
The best software is invisible.
People want problems solved, not another app to learn. Your team already has agents they like. Kody should make those agents better without asking anyone to open something new.
In the product: There is no Kody chat app. Your agent reaches Kody over MCP, and you keep working where you already work. What is Kody?
Your agent is your choice.
Agents change fast, and people on the same team prefer different ones. Switching agents should not mean starting over, and one person’s pick should not decide everyone else’s.
In the product: Every agent you connect shares one account: the same memories, credentials, packages, and jobs. Org accounts let a whole team share that home. Connect your agent
Keys belong out of the prompt.
An agent can be talked into anything that is in its context. A credential pasted into chat or a .env file is one bad instruction away from leaking.
In the product: No capability returns a credential, and there is deliberately no plaintext secret_get. Code uses keys by name, and Kody fills them in at the network boundary, only for hosts you approved. Secrets
Work done once should stay done.
Asking an agent to redo the same task every morning spends tokens and invites drift. Once something works, it should keep working without a model deciding how to do it again.
In the product: Jobs, workflows, webhooks, and email run saved packages with no model in the loop. Kody runs no chat-model loop and bills no chat tokens. Jobs, workflows, and webhooks
You own what your agents build.
The useful things your agents make should not be locked inside one vendor’s chat history or someone else’s runtime.
In the product: Packages are code in your account that you can open, change, schedule, and republish. Installing a public package gives you a fork you own. Kody itself is Fair Source. Sharing a package