Credits & usage

Your subscription includes a monthly token allowance on MiniMax M2.7 — the platform’s default model. That allowance pays for the actual LLM usage your agent generates. We track your usage against the token price of M2.7, so the same dollar on your plan buys more as model prices drop over time.

What counts as credit usage

• Every message your agent processes uses some tokens (input + output)
• Web searches via Serper cost ~$0.001 per search
• Sub-agents spawned for parallel work each use their own credits
• Scheduled task runs use credits just like manual runs
• Vision analysis on images uses a bit more due to the vision model

What does NOT count:

• Browsing the web — your agent’s built-in browser (Camoufox sandbox and the Browser Control extension) is free. Load as many pages as you want
• Reading your own memory files (MEMORY.md, USER.md)
• Accessing your workspace files (no credit cost to read/write local files)
• IMAP/CalDAV sync (your email/calendar providers are free to us)
• Telegram/Discord/Slack message delivery (platform APIs are free to us)
• Stripe billing activity (no credits consumed)
• UI interactions (clicking around the dashboard)

Your included credits

Tier	Monthly included tokens
Trial	~5M MiniMax M2.7 tokens (for the 15-day trial period)
Standard	~5M MiniMax M2.7 tokens/month
Pro	~15M MiniMax M2.7 tokens/month

These refill on your billing renewal date. Included tokens are priced against MiniMax M2.7 — if you use a different model through BYOK, your agent pays that model’s bill directly to the provider and doesn’t touch this allowance at all.

Checking your usage

Dashboard → Usage shows:

• Current month’s usage — how much of your included credits you’ve used
• Breakdown by source — chat, scheduled tasks, web search, vision, sub-agents
• Most expensive operations — so you can spot unexpected spikes
• Historical trend — your usage over the past 3 months
• Projected end-of-month burn — based on current pace, where you’ll land

The dashboard also shows your pace visually: a simple bar going from 0% (freshly refilled) to 100% (credits gone).

Warnings and alerts

We proactively warn you as your credits deplete:

• At 80% usage — your agent pushes a warning through your notification channels and flags a credit-alert banner on the dashboard.
• At 100% usage — your agent pushes an “exhausted” notification and activity is throttled (see below).

What happens when you hit 100%

Your agent does NOT stop working abruptly. Instead:

1. You’re notified through your notification channels and a credit-alert banner appears on the dashboard.
2. New action requests are throttled — your agent stops starting new expensive work (deep research, sub-agent spawns, heavy scheduled tasks) until credits are topped up.
3. Direct conversation with you keeps working — you can still chat, read memory, and ask quick questions so you’re not locked out of your own agent.
4. You’re prompted to act — buy a credit pack, upgrade your plan, enable BYOK, or wait for next month.

You can choose to:

• Buy a credit pack — immediate top-up, continues where you left off
• Upgrade to a higher tier — higher monthly allowance + prorated billing
• Enable BYOK — route through your own provider key and bypass our included pool entirely
• Wait — do nothing until next month’s renewal

Credit packs

Available from the Credit packs card inside Settings → Billing:

Pack size	Price	Approx MiniMax M2.7 tokens
~4.5M tokens	$5	4.5M tokens loaded
~9.5M tokens	$10	9.5M tokens loaded
~24M tokens	$25	24M tokens loaded
~48M tokens	$50	48M tokens loaded
~97M tokens	$100	97M tokens loaded

Credit packs:

• Never expire — unused tokens roll over indefinitely
• Are used AFTER your monthly allowance — so your monthly allowance doesn’t get wasted
• Are shared across all categories — a single pack can be spent on chat, search, vision, sub-agents, or anything else

BYOK — bring your own keys (available today, every tier)

Premium models (Claude, GPT-4o, Gemini, DeepSeek, OpenRouter, self-hosted) are available only through BYOK. We don’t resell premium models against your included credits — if you want to run on a model other than MiniMax M2.7, plug in your own API key and your agent uses it directly.

Open the Bring Your Own Key card in Settings → Agent and add a key for any supported provider:

• Anthropic (Claude)
• OpenAI (GPT-4o, GPT-4o-mini, o-series)
• Google (Gemini)
• MiniMax (direct, including their flat-rate coding plan)
• DeepSeek
• OpenRouter (one key, hundreds of models)
• Any OpenAI-compatible endpoint — point it at your own relay or self-hosted model

When a BYOK key is present for a provider, your agent uses it directly and we don’t charge you any token credits for requests that route through it. You keep paying the provider on their bill. This is the cheapest path for heavy users — especially if you grab a MiniMax M2.7 flat-rate coding plan and drop that key in, which effectively gives you unlimited model usage for a fixed monthly fee.

BYOK is available on every tier, including trial. It is not a Pro-only feature — if you need Claude or GPT-4o on a Standard plan, that’s supported today.

Tips for staying within budget

1. Pause monitoring you don’t read. A daily competitor monitor that delivers a digest nobody looks at is still burning credits. Pause it.
2. Batch your research. Asking 10 questions in one conversation is cheaper than 10 separate conversations (context is shared).
3. Scope scheduled tasks tightly. “Summarize today’s inbox” is cheap. “Summarize today’s inbox, analyze my response patterns, compare to last week’s, and suggest improvements” is 5× more expensive.

The credit bar is there for a reason

If you see your credit usage climbing faster than you expected, open Dashboard → Usage and look at the breakdown. 90% of unexpected spikes trace back to one scheduled task that’s doing more than you meant it to.