CloudCode.ONE in 15 seconds: no subscription, keep your tools, 1M-token context
We made a 15-second film. It says three things about CloudCode.ONE and then gets out of the way, so here it is, followed by the longer version of each line.
No subscription.
There's no monthly fee and no plan to pick. You buy credit when you need it — $2, $5, $10 or $20 at a time — and pay only for the tokens you use. Credit never expires, so a quiet month costs nothing, and nothing burns down at the end of it.
That makes the cost follow the work. A weekend project, a one-off refactor and a month of heavy agent sessions each cost what they use. Here's what a million tokens costs on a few of the models today, in USD:
| Model |
Input / 1M |
Cached / 1M |
Output / 1M |
glm-5.3 |
$1.26 |
$0.234 |
$3.96 |
glm-5.3-flash |
$0.165 |
$0.033 |
$0.55 |
deepseek-v4.1-flash |
$0.27 |
$0.0054 |
$1.08 |
mistral-large-4 |
$1.54 |
$0.154 |
$4.62 |
The front page always has the full, current list.
Keep your tools.
CloudCode.ONE doesn't ask you to switch editors or learn a new agent. It speaks both of the APIs coding tools already use, Anthropic's and OpenAI's, so you change a base URL and an API key and keep everything else:
- Claude Code: set
ANTHROPIC_BASE_URL to https://api.cloudcode.one and use your CloudCode.ONE key as the auth token. Your settings, hooks, skills and MCP servers stay exactly as they are.
- Cline, Roo Code, Continue, OpenCode and anything else that speaks the OpenAI API: choose "OpenAI Compatible" and set the base URL to
https://api.cloudcode.one/v1.
- Harness, our free desktop agent for Windows and Ubuntu, if you'd rather not configure anything: sign in with your browser and it connects to your account by itself.
One key works in all of them. Once you're signed in, the setup guide has copy-paste settings for each tool with your key already filled in.
1M-token context.
Most of the models here read up to a million tokens at once: GLM-5.3 and GLM-5.3-Flash, DeepSeek V4 Pro and V4.1 Flash, and Mistral Large 4. That's room for a whole repository, a long log, or an agent session that runs all afternoon without losing track of what it did in the morning.
A long context is only useful if you can afford to fill it, and an agent resends the whole conversation on every turn. So the gateway keeps each account's context warm in the cache, and cached input bills at a fraction of the normal rate — $0.234 instead of $1.26 per million on GLM-5.3, $0.154 instead of $1.54 on Mistral Large 4. Long sessions run mostly from cache, with nothing to configure.
One tip for Claude Code: add [1m] to the model name, as in glm-5.3[1m], so Claude Code plans for the full window instead of its default 200K. It strips the suffix before sending, so the gateway sees plain glm-5.3.
Your agentic coding partner. Available now.
That's the whole pitch: pay for what you use, keep the tools you know, and give your agent room to think about the whole codebase.
New here? Create an account, top up from $2 or redeem a coupon, and your key is ready straight away.