Grok Build Pricing and Access: Plans, API Cost and Parallel Agents

The plans at a glance

Grok Build is free to try, and paid plans raise the limits. Since October 12, 2026, The X Subscription is available to new customers. This is the lineup as of October 31, 2026:

Payment is not required for evaluation. The paid ladder is now a set of multiples of Plus: Premium gives 4x, Super 15x and Ultra 50x, and xAI says each plan includes at least twice as much coding usage with its models as the comparable Cursor plan. Usage beyond the plan is billed at API list prices through top-ups. The announcement does not break the shared pool out per product, so it does not say how much of it Grok Build alone can use.

What is published, and what is not

This guide gives the prices xAI publishes. It deliberately does not invent a request quota because xAI describes limits relatively and does not publish one stable request count for every Grok Build tier.

The practical planning advice still transfers: your burn rate depends on how many agents you run in parallel because each agent independently reads files, runs tools and reasons. Check the account usage view for your current limits instead of budgeting from an old article.

If you are pricing a parallel-agent workflow, that is the number that decides your tier, and it is the argument developed in the Grok Build agent swarm guide.

The pay-per-token route: the xAI API

For unattended work, XAI_API_KEY is the headless path and swaps account limits for per-token billing. Grok 4.7 is the default model of Grok Build. Its standard API rate is $2 per million input tokens, $0.50 per million cached input tokens and $6 per million output tokens, rising to $4, $1 and $12 when a prompt reaches 200K tokens.

Grok 4.7 has a 500K context window. Its $2 input rate matches Claude Sonnet 5.5, while its $6 output rate is lower than Sonnet's $10. Grok 4.7 Fast is available only in Cursor and Grok Build, billed through your plan and not on the free tier. Compare the current figures with the Claude Code rates before choosing a CI model.

The API key keeps working while credit and rate limits allow it, so unattended usage needs a budget and alerts. Set a spend cap on the xAI side and keep the key in your CI secret store rather than in the shell you code in.

# Interactive: browser login on first launch
grok

# Headless / CI
export XAI_API_KEY=...
grok -p "run the migration and report what changed"

What running several Grok Build sessions at once does to the cost

On a plan, extra sessions add no fee. They consume the same account allowance faster, because each agent independently reads files, runs tools and reasons. Four Grok Build terminals draw from exactly the allowance four separate terminals would, which is why the deciding number is how many agents you keep busy rather than which app they run in.

On an API key the effect is a direct cost instead of an earlier wall. Grok 4.7 bills $2 per million input tokens, $0.50 per million cached input tokens and $6 per million output tokens at the standard context tier, with higher pricing above 200K context, so spend over an hour tracks how many agents are actually working.

How to divide work between those sessions, and how to stop them duplicating each other, is in the Grok Build agent swarm guide.

What CodeAgentSwarm costs on top

Nothing on the model side. CodeAgentSwarm does not resell Grok tokens and does not sit between you and xAI. You pay xAI or X for model access, exactly as you would running grok in a bare terminal.

Running four Grok Build terminals inside CodeAgentSwarm therefore adds no swarm fee on the xAI side. It draws from the same allowance four separate terminals would, just with the sessions labelled, their status visible and their history searchable.

What it does add is usage visibility: the app shows your xAI usage alongside your other agents, so you can react before a session stalls.

How Grok Build pricing compares

Set against the other CLIs, Grok Build starts free and its standard API tier costs $2 per million input tokens, $0.50 for cached input and $6 for output.

• Evaluation starts at $0. The official Grok Build page says it is available to try free.
• Paid plans start at $8. X Subscription Plus is below the $20 entry plan for Claude Code, and paid access is optional for a first test.
• Premium costs $30, Super $100 and Ultra $200. They give 4x, 15x and 50x Plus usage from a pool shared with Grok and Cursor.
• The current API is $2 input and $6 output. Cached input is $0.50 per million tokens, and requests above 200K context cost more.

The practical reading: use the free access to test Grok Build, pay for higher limits only when your account usage proves you need them, and use the API for controlled automation. Running it alongside another agent also gives you separate provider limits, which is the argument in the multi-CLI swarm guide.

Every number in that table has its own page with the full tier list and the fine print: Claude Code, Codex, Kimi Code, OpenCode, Antigravity and Cursor CLI.

How to verify on your machine

# Which CLI do you actually have?
grok --version

# First interactive launch surfaces auth if you are not signed in
grok

Grok Build keeps its data under ~/.grok, relocatable with GROK_HOME. That is also where sessions live, which matters when you want to find an old one: see the Grok Build conversation history guide.