Token Usage & Budgets

Codewick uses a token-based usage model. Every time an AI model processes your request, it consumes tokens — and those tokens count against your monthly budget. This page explains what tokens are, what uses them, and how to manage your budget effectively.

What are tokens?

Tokens are the units AI models use to process text and code. A token is roughly 3-4 characters of English text, or about 0.75 words. Code tends to tokenize less efficiently than prose — special characters, indentation, and syntax all consume tokens.

There are two types:

• Input tokens — Everything sent to the model: your message, conversation history, code context, file contents, and system instructions.
• Output tokens — Everything the model generates: code, explanations, plans, and review findings.

Output tokens generally cost more than input tokens. A task that generates a lot of code (like scaffolding a new feature) uses more output tokens. A task that analyzes existing code (like a review) uses more input tokens.

Why token counts vary

The same question can consume very different amounts of tokens depending on:

• Conversation length — Longer conversations include more history as input context.
• Number of files in context — More referenced files means more input tokens.
• Pipeline stages activated — Each stage is a separate model call with its own token usage.
• Response complexity — A multi-file implementation uses more output tokens than a one-line fix.

What consumes tokens

Any action that triggers one or more pipeline stages consumes tokens:

• Chat messages — Sending a message in the chat panel triggers the pipeline. More stages activated means more tokens used.
• Debugging runs — When Codewick’s debugging stage analyzes errors, it reads code context and produces diagnostic output.
• Review passes — The review stage reads generated code and produces quality assessments.
• Project creation — Using Codewick to scaffold a new project runs orchestration and planning stages, which consume tokens.
• Follow-up questions — Each follow-up message in a conversation includes prior context, so token usage grows as conversations get longer.
• Retries — If a pipeline stage fails and retries automatically, the retry consumes additional tokens.

What does NOT consume tokens

Many Codewick features work entirely locally and never touch an AI model:

• Editing code in the editor (typing, pasting, find-and-replace)
• Terminal commands and output
• File explorer operations (creating, renaming, moving files)
• Git operations (commits, branches, diffs, push, pull)
• Built-in browser preview
• Restoring checkpoints from the timeline
• Navigating the interface, changing settings, switching files
• Searching within files or across your project
• Installing extensions or configuring preferences

Tip

If you’re on a tight budget, do as much manual work as you can in the editor and terminal, and use AI for the parts where it adds the most value — planning complex features, debugging tricky issues, and reviewing code quality.

Monthly token budgets

Each subscription tier includes a monthly token budget that resets at the start of your billing period.

Tier	Monthly budget	Target user
Lite	Lower budget	Casual use, learning, small projects
Daily	Moderate budget	Regular daily development
Pro	Large budget	Full-time professional development
Max	Highest budget	Teams and power users

Note

Exact budget amounts are being calibrated during beta. Current subscribers receive generous allocations while we finalize pricing. Check your workspace settings for your current budget and usage.

How budgets reset

Your budget resets on the same day each month, aligned with your subscription start date. Unused tokens do not roll over — each billing period starts fresh. The reset date is visible in the usage panel.

Live usage counter

Your current token usage is always visible in the workspace. Look for the usage counter in the bottom status bar — it shows tokens used and an estimated dollar cost for the current billing period.

The counter updates in real time as pipeline stages complete. You can watch it increment during a task to get a sense of how much each stage costs.

Expanded breakdown

Click the usage counter to expand a detailed view showing:

• Per-stage breakdown — How many tokens each pipeline stage consumed across all your tasks this period.
• Input vs. output split — The ratio of tokens sent to models versus tokens generated by models.
• Daily trend — A simple chart of your daily token consumption over the current billing period.
• Projected usage — An estimate of whether you’re on track to stay within your monthly budget based on your current pace.
• Top tasks — The individual tasks that consumed the most tokens, so you can identify expensive operations.

This breakdown helps you understand where your tokens are going. If Building and UI Generation are dominating, your tasks are code-heavy. If Orchestration and Review are high, you’re running many small tasks.

Session spend limit

You can set a per-session spend limit to prevent any single work session from consuming too much of your budget.

1. Open Settings > Usage.
2. Set a token or dollar limit under Session spend limit.
3. Codewick will warn you when you approach the limit and stop AI features when you hit it.

This is useful if you want to timebox how much budget a particular task or exploration uses. When the session limit is reached, you can start a new session to continue working (the monthly budget still applies).

Tip

Session limits are especially useful when experimenting. Set a low limit before exploring a new approach so you don’t burn through your budget if the direction turns out to be wrong.

What happens when you reach your limit

When you hit your monthly budget:

• AI features are disabled. Chat messages won’t trigger the pipeline. You’ll see a clear indicator that you’ve reached your limit.
• Non-AI features continue working. The editor, terminal, file explorer, git tools, and browser all work normally. Your project is never locked.
• You’ll see your reset date. The usage panel shows exactly when your budget resets.
• Upgrade prompt. If you’re on a lower tier, you’ll see an option to upgrade for a higher budget.

You’re never charged beyond your tier’s budget. There are no surprise overages.

Usage meter

The status bar displays a usage meter — a percentage showing how much of your monthly budget you’ve used. It includes a plain-language label so you can gauge your usage at a glance:

Percentage	Label	Status bar color
0-50%	On track	Green
51-75%	Moderate use	Green
76-90%	Heavy use	Amber
91-99%	Near limit	Red
100%	Limit reached	Red

The color transition is gradual, not sudden — you’ll notice it shifting as you move between ranges.

Token top-ups

Note

Token top-ups will be available after the beta period ends. During beta, your tier’s monthly budget is your total allocation. We’ll announce top-up pricing and availability when the feature launches.

After beta, if you exhaust your monthly budget before the reset date, you’ll be able to purchase additional tokens without upgrading your tier. Top-ups are one-time purchases that expire at the end of your billing period. They’re designed for months when you have unusually heavy usage, not as a regular supplement.

How priority mode affects token cost

Your priority mode has a direct impact on token cost:

• Cost mode routes to cheaper models, so each token costs less. Your budget stretches further.
• Quality mode routes to premium models with higher per-token prices. You get better output but use your budget faster.
• Speed mode tends to use mid-priced models that respond quickly.
• Balanced mode lands in the middle, optimizing for reasonable cost at good quality.

If you’re running low on budget, switching to Cost mode for the remainder of the period is an effective way to stretch your remaining tokens.

Tips for managing token usage

• Be specific in your messages. Vague requests cause the orchestrator to explore more possibilities, using more tokens. “Add a submit button to the login form that calls the /auth endpoint” is cheaper than “improve the login page.”

• Start new conversations for unrelated tasks. Long conversations accumulate context that inflates input tokens on every message. A fresh conversation starts with a clean context window.

• Use Cost priority mode for exploration and learning. Switch to Quality mode when you’re ready to commit to a solution.

• Review the per-stage breakdown periodically. If one stage is consuming a disproportionate amount, consider whether you can adjust your workflow.

• Pin only what’s needed in context. Excess pinned files increase input tokens on every pipeline run. See Context Management for details.

• Break large tasks into smaller ones. A single massive request activates all stages with heavy context. Several focused requests can be more token-efficient overall.

• Use @ mentions instead of describing files. Saying “refactor @utils/auth.ts” is more token-efficient than pasting code into the chat or describing the file’s location.