Plans and limits
Every plan of Context Mode Gateway has every feature, in Claude Code and Codex. Start free: 2,000 requests, every feature, no card. When your requests run out, nothing breaks: your requests go straight to the model.
Key facts
- What it is
- What each plan includes, what counts as a request, and what happens when your requests run out.
- Who it is for
- Anyone who connects Claude Code or Codex to the gateway, alone or as a team.
- Clients
- Claude Code and Codex, on every plan. Your own Claude or OpenAI account pays the model, and we add no fee on tokens.
- Measured result
- Not measured: this page states plan terms, not results. For cost, see Benchmarks.
- Limits
- The free requests are given once per account, not every month. Prices are not on this site: the console shows each price before you pay.
The plans
| Plan | Requests | Features | When the requests run out |
|---|---|---|---|
| Free | 2,000, once per account | Every feature. No card. | Paused until you move to Pro |
| Pro | 50,000 a month | Every feature | Paused until next month, or add a request pack |
| Team | 50,000 a month for each seat, in one pool | Every feature, one Cage policy for the org, one bill | Paused for the team until next month, or add a request pack |
| Enterprise | Set in the contract | Every feature, on our Cloudflare account or in yours | Paused, never cut off |
Every feature means Context Saving, Thinking in Code, Cage, Memory, Skills, Search, Session resume, Context limit, Reply style, Live and Analytics, in Claude Code and Codex.
Free
- 2,000 requests, given once when you sign up. They do not come back each month.
- Every feature, in Claude Code and Codex.
- No card.
To start, run npx @context-mode/cli, or sign in to the console. See the quick start.
Pro
- Every feature, in Claude Code and Codex.
- 50,000 requests a month. The count starts again each month, on a yearly plan too.
- Request packs when you need more.
- Pay by the month or by the year.
Request packs
On Pro and Team, a request pack adds 25,000 requests. You buy it in the console and pay first. We never charge you after the fact. A pack's requests are used only after the month's included requests, and they last 12 months.
Auto-reload is off until you turn it on. When it is on, it buys a new pack when your balance runs low, and it stops at a monthly limit you set.
Team and Enterprise
Team gives each seat everything in Pro. The team shares one pool of requests, 50,000 a month for each seat, and its request packs. Admins set one Cage policy for the org, and members cannot loosen it. One bill goes to the team owner. A team has at least 2 seats. Each member's memory stays their own.
Enterprise comes three ways, on a yearly contract:
- Enterprise Cloud: our gateway, with contract terms.
- Enterprise Dedicated: your own gateway, with its own hostname and secrets, in our Cloudflare account.
- In your Cloudflare account: the same gateway, deployed in your account. You hold the data, the keys, the logs and the Cloudflare bill.
There is no on-premises build. For Team or Enterprise, write to us on Discord.
When your requests run out
Nothing breaks. No request fails because of your plan: not on Free, not at the limit, not when a card fails. Here is what happens.
- A running session keeps going as it was. A session that started with every feature on keeps them until it ends, within a limit, so your agent's tools do not vanish in the middle of a task.
- Your next new session starts paused. Its requests go straight to the model, with your own credential. Your agent keeps working, with no error and no refused request.
- These pause: Context Saving, Thinking in Code, Memory and Skills. The savings ledger stops, and Live shows one row for each request, marked paused.
- These stay on: Cage core protections. Calls the gateway rewrote earlier in a conversation are put back before the request goes to the model, so the model reads a consistent history. Your stored memory, skills, Cage rules and archive are kept.
- It comes back on when you move to Pro, buy a request pack, or your next month starts on Pro or Team. A paused session switches back at its next request. On Free, the free requests do not come back: move to Pro to resume.
If we cannot read your plan or your count, you get every feature. If a card payment fails, Pro stays on while the payment is tried again. A problem on our side never stops your agent.
How you are told
Once, clearly, and not on every turn.
- At 80% of your requests: one line in your terminal and one e-mail, once in the month. On Free, once.
- When the pause starts: one line in your terminal and one e-mail. It says that your requests now go straight to the model, what paused, what stays on, and where to resume.
- While paused: a short reminder, at most once per new session and at most once a day. Never on every turn, and never on a subagent's turn.
- In the console: a usage meter on Overview and Billing, and while paused, a bar at the top of each page that you can hide for a day.
- When it is back on: one line in your terminal.
Refunds
Within 14 days of your first Pro payment, or of buying a request pack, you can cancel and get back the part you did not use. The used part is the larger of the time used and the requests used, so you get back only what you did not use. This works once per customer for each product.
The Billing page shows the exact amount, with the days and the requests used, before you confirm. Renewals, later packs and requests after 14 days are not refunded. When you cancel, your plan stops renewing and runs to the end of the period you paid for.
Fair billing promise
We never block your agent. We never charge you for usage you didn't buy. Cancel in two clicks. Changed your mind in the first 14 days? You get back what you didn't use.
FAQ
Is there a free plan?
Yes. Start free: 2,000 requests, every feature, no card. The 2,000 requests are given once per account; they do not come back each month.
What counts as a request?
Each model request your agent sends through the gateway, subagents included. A request the gateway retries counts once. Token counts, model lists, console reads and the gateway's own side calls do not count, and neither does a request the provider refuses for a reason we did not cause.
What happens when I run out?
Nothing breaks. Your requests go straight to the model, so your agent keeps working. Context Saving, Memory, Skills and Thinking in Code pause, and Cage core protections stay on. See step by step.
Do I need a card to start?
No. Free needs no card. You add one only when you move to Pro or buy a request pack.
How do I cancel?
In two clicks, from the Billing page in the console. Your plan runs to the end of the period you paid for.