What Tokenaissance is

Tokenaissance is a hosted coding agent. You describe a task, it works in a container with four tools — write, read, edit, bash — and you watch it work: the answer streams in, each tool call appears as it is made, and every reply carries what it cost.

The difference worth caring about is where the conversation lives. Most chat products keep a transcript in the browser and hope the connection holds. Tokenaissance keeps it on the server, next to the sandbox that does the work, and treats the browser as a viewer: reload the page and the conversation is where you left it.

What you can ask for

• Write and change files. "Create a worker.ts that caches responses for an hour." The agent writes it, then reads it back to check what it wrote.
• Run commands. npm install, a test suite, a one-off curl — output streams while the command runs, so a slow step looks slow instead of frozen.
• Investigate. "Why does this build fail?" It runs the build, reads the error, opens the file, fixes it, runs it again.

The container has outbound network, so installing a package or fetching a doc page works. Each user gets their own container; nothing of yours is mounted into it.

What it shows while it works

A conversation is not just text:

| On screen | What it means | | ------------------------ | ---------------------------------------------------------------------------------------------- | | Tool cards, in place | The call is made where the model made it, and the card shows running… until the result lands | | Thought process | The model's reasoning, folded away by default, expandable when you want to check it | | 9,628 tokens · $0.0017 | What that single answer cost — the number comes from the provider, not an estimate |

What it costs

One credit per prompt. New accounts start with ten, so you can do ten real tasks before paying anything. The prompt is charged when it is accepted; if you run out, the conversation says so and links to the purchase page instead of failing silently.

What it deliberately does not do

• It does not promise deletion. Removing a conversation marks it removed and filters it out of your list; the rows stay in the object's storage. We say "removed", not "deleted", because that is what happens.
• It does not bill by token. A prompt costs one credit regardless of length. Token accounting is exact, but settlement happens per prompt.
• It does not hide a failed provider behind a circuit breaker. If the model call fails, you get the error.

Ready to try it? Start a conversation and ask for something concrete.