03 · Sign-in, accounts, credits
Two things people skip that directly set your capability ceiling and your monthly bill: which identity you sign in with, and how credits actually burn.
Personal or enterprise
The sign-in page asks you to choose. That choice decides which features you can see.
| Personal | Enterprise | |
|---|---|---|
| For | Your own plan, or the free tier | Your company bought seats and assigned you one |
| Credits | Your own allowance | Company allowance, shared across the team |
| Projects | Yes, member count capped by plan | Yes, higher member caps |
| Enterprise Agent platform | No | Yes — cloud Agent runtime |
| Enterprise knowledge base | No | Can link custom knowledge bases configured in the admin console |
| Central controls | No | SSO, IP allowlist, per-department credit caps, audit |
| Agent Mail | Yes, personal instance | Yes, enterprise instance, isolated from personal |
If your company bought enterprise, sign in as enterprise
The enterprise credit pool is shared across the team and usually far more generous than a personal plan, and the Enterprise Agent platform is only visible to an enterprise identity.
Note that the two identities keep separate Skill libraries. Do not try to work across both.
What credits are
WorkBuddy uses credits to meter AI resource consumption. Every conversation-related request costs credits, and the amount depends on two variables:
Model token pricing. Stronger models reason better and know more, but their input and output tokens cost more to process. The same volume of tokens costs far more on a flagship model than on a fast one.
Task complexity. A short question costs little. Long code analysis, multi-turn conversations and knowledge-base retrieval need a lot more reasoning and token handling, and the cost scales accordingly.
The credit pool is shared with CodeBuddy
Same account, same credits, no separate subscription.
What burns credits fastest
Roughly ordered:
| Burn rate | What you did | Why |
|---|---|---|
| Very high | Summoning an expert team | Several experts run in parallel over multiple turns — officially 3–5× a single expert |
| Very high | Long conversations with Max mode on | Context is never compressed, so every turn resends the full history |
| High | Context window set to 1M | Every turn is billed against the big window; a bigger window does not make the model smarter |
| High | Complex task plus complex Skill | Many tool calls, many correction rounds |
| Medium | Summarising data a connector fetched | Connector reads and writes are free; understanding and summarising them is not |
| Low | Short questions, format conversion, copy polishing | Single turn, short context |
| Free | Memory extraction | Officially at WorkBuddy's own expense |
| Free | Browsing Inspiration, reading mail in the Agent Mail UI | Never enters model context |
| Free | Library storage, organising, searching, manual editing | Only AI processing of that content is billed |
Three habits that keep the bill down
- Leave the context window at 200K and Max mode off. Raise it only for genuinely long documents.
- One task per conversation. A conversation that runs too long makes you pay for irrelevant history on every turn. When it drifts, start a new task and restate the background.
- Prove the direction on a cheap tier first. Validate a new kind of task in fast mode, then switch to maximum for the final run.
How deduction works
The system spends the credits that expire soonest first. Where expiry dates match, base credits go before bonus credits.
Personal credits are issued monthly, valid that month, and do not roll over; bonus credits expire one month from the date they land. Anything unused at month end is gone, so use it.
If your paid plan expires while a top-up pack still has credits, you can keep using the pack until it expires or runs out.
Plan tiers
Four personal tiers since 1 July 2026, when the single "professional" plan expanded into standard / advanced / ultimate:
| Free | Standard | Advanced | Ultimate | |
|---|---|---|---|---|
| Monthly | Free | ¥99 | ¥199 | ¥999 |
| Credits per month (promo) | 500 | 4,000 | 9,000 | 50,000 |
| Automated tasks | 3 | 15 | 30 | 99 |
| Model selection | Auto routing | All models | All models | All models |
| Projects | 5 | 10 | 15 | 20 |
| Personal assistants | 3 | 5 | 8 | 10 |
Monthly auto-renew is 30% off, annual is 30% off, annual auto-renew takes a further 20% off. Enterprise plans start at ¥198 per seat per month (SaaS) and ¥316 per seat per month for the private-cloud edition with a 100-seat minimum.
Full pricing, top-up packs, legacy-user transition and refund rules are in Credits & plans.
Run the free tier for a week first
500 credits a month is enough to work through every task in this book's Get started chapter, and enough to judge whether the product is what you want.
Decide on a tier afterwards. That beats subscribing on day one — and note that using any credits from an order voids its refund eligibility.
Checking usage
Sign in at workbuddy.cn → avatar (top right) → personal home → plan and usage. You will see your current subscription plus available and used credits per resource pack.
Enterprise members also see the allowance their company assigned them.
Credits missing or consumption looks wrong
Check the balance and usage detail under personal home → usage management first. If the numbers genuinely do not add up, email a screenshot to workbuddy@tencent.com and the team will look into it.
Next
Account sorted. Go learn the window.