If your agent hits the 5-hour limit before your task is finished, you either stop, or you top up your wallet for a few dollars and keep going right now. Starter + wallet is the clean move.
中文版本: /plans-and-credits.cn
Your plan gives you a dollar allowance per 5-hour window. If the task is small, no problem. If the agent is doing a big refactor, you will run out in the middle. Two things can happen then:
Fikra Pro, running a long agent refactor.
Wallet credit lives in the same wallet as pay-as-you-go. Credit does not expire.
| Top-up | USD | Roughly covers at $0.1875/M input | At $0.75/M output |
|---|---|---|---|
| $5 | $5 | 26.6M tokens input | 6.6M tokens output |
| $10 | $10 | 53.3M | 13.3M |
| $25 | $25 | 133M | 33.3M |
| $100 | $100 | 533M | 133M |
Input token share of DeepSeek V4.1 Flash 1017 will differ based on cached hit ratio. Figures rounded to one decimal.
Pro plan for one key, $5 spare credit. Real agentic workload coverage without watching the clock.
Max plan across 3-5 seats; $25 spare credit per seat. Floater coverage of an entire sprint.
Mac plan - $60 per month - with a $100 top-up for agentic PRs running 24/7.
/account/signup./plans or stay on pay-as-you-go./pricing, pick a top-up card ($5, $25, $100 or custom).All credits stay on the wallet and burn at list prices on any model selected, including the DeepSeek V4.1 lanes that are popular in China.
Plans list directly from /models. Every model Fikra adds to the network is included in your plan the moment it ships, no separate entitlement, no extra opt-in - your allowance now applies to the new model automatically.
New models available in all plans, including ones Fikra adds zero-day after lab releases. Want to try a brand-new release right now? Ask your agent to call /v1/models the same hour it hits the wire.