Questions
before you connect.
Straight answers about what Kody is — and what it is not.
Does Kody replace Claude, Cursor, ChatGPT, or Codex?
No. Kody is not a chat app and does not replace those products. You connect the agent you already use — Claude, ChatGPT, Cursor, Codex, or any other MCP-capable host — to your Kody account.
That agent gains durable capabilities that outlive the conversation: memory, credentials, saved code, schedules, and execution. You keep talking to the agent you already pay for.
Does Kody call a model / make inference calls?
Kody does not run its own agent loop and does not bill for chat tokens. Your connected agent does the thinking. Kody holds the result so the work can run again without another model in the loop.
Search indexing uses a small embedding model. That is not a chat model and is not billed as inference.
What is a package vs a prompt vs a memory?
Kody does not treat prompts as a saved artifact. A prompt is what you type into your agent.
A package is reusable saved code your agent writes and improves over time. Unlike a one-off execute, it lives in a repo and can expose exports, scheduled jobs, inbound webhooks, and even a small web app.
A memory is a durable fact or preference Kody keeps about you. Memories follow the account across every connected agent. Facts saved only in a host — a Claude Project, or Cursor rules that are not also Kody memories — stay invisible to your other agents.
Can my team share one Kody account?
No. Isolation is the feature. Each signed-in user gets a fully isolated assistant: their own packages, jobs, secrets, memories, and storage. There is no org tenancy or shared-team workspace.
Your assistant is yours unless you publish a community package. Publishing puts a snapshot in the public catalog; installing creates a fork the installer owns.
How do secrets work? Can my agent read the key?
Your agent cannot read the key. No capability returns a secret value — there is deliberately no secret_get.
Code and the agent refer to secrets by name. Kody substitutes them at the network boundary, only for hosts you approved. Use secrets so your agent can call the services you already use without pasting keys into the conversation.
What happens if I switch agents?
Packages, secrets, jobs, memories, and the rest of your assistant stay on the account. Connect the new host from Get started. The same assistant is available from every MCP-capable agent you connect.
Host-only context — a Claude Project, local Cursor rules, or notes that never became Kody memories — does not transfer.
What does the free plan include?
Every plan is the whole factory. The free plan is $0 with volume caps: 10 saved packages, 5 scheduled jobs (no faster than every 15 minutes), 250 execute calls per day, and the rest of the entitlements on Pricing. Paid plans raise the caps.
Can I export or self-host?
You can download a JSON export of your account from Account settings. Secret values are never included; secret entries export metadata such as names, hosts, and allowlists only.
Kody is Fair Source. The source is on GitHub: you can read it, fork it, and run your own deployment.
How is this different from a Claude Project, a local folder of skills, or n8n?
A Claude Project keeps context inside one host. Kody memories, packages, and secrets follow the account, and scheduled jobs run in the cloud with no model in the loop.
A local folder of skills lives on one machine. Kody packages and jobs run on Cloudflare Workers whether or not your computer is on.
n8n is a workflow builder you assemble by hand. Kody is the home your agents share: your agent writes the code, saves it as a package you own, and can schedule it so the work keeps running without another inference call. Installing a community package puts a fork in your account, on your credentials.
How do I get started?
Read What is Kody? — it needs no account. Then open Get started, pick the agent you want to connect, and complete OAuth.
This deployment may require an invite. Without a code you can join the waiting list from Sign up.