Disclosure: some links on this page may be affiliate links; we may earn a commission at no extra cost to you. Read the disclosure.

By the toolist.ai editors. Researched from Claude Code's own pages and hands-on tests published by others, not hands-on tested by toolist.ai. Updated October 13, 2026.

Claude Code review: pricing, features and what testers found

Claude Code is Anthropic's coding agent: it works through your codebase, changes files, runs commands and tests, and opens pull requests from your terminal, IDE, desktop app, the web or Slack. There is no free plan: it comes with Claude Pro at $20 a month (or $200 a year) and the higher Claude plans, or runs on pay-as-you-go API tokens.

Our take

Claude Code suits developers, and non-developers willing to learn as they go, who want to hand an agent whole tasks such as bug fixes, tests and multi-file refactors: it comes with Claude Pro at $20 a month, with Max at $100 or $200 a month for 5 or 20 times Pro's per-session usage, or runs on pay-as-you-go API tokens. Zapier's Miguel Rebelo rates its delegation-first harness ahead of Cursor's for multi-agent orchestration and CI/CD and tells non-developers to start with it, while developers who want to stay close to the code should pick Cursor. In The New Stack's 2026 tests, Jessica Wachtel got a working CLI tool with 62 passing tests in about seven minutes from parallel agents on a Max plan, but found Claude Code's memory does not carry from one repo to another and each of its sessions cost at least twice as much as Grok Build's; Tom's Guide's Amanda Caswell preferred OpenAI's Codex for building a real application. The main catch is usage: Claude Code shares a five-hour session limit and a weekly limit with Claude chat, Anthropic publishes no fixed numbers, and past them you wait for the reset, upgrade, or pay API rates through usage credits. On Pro and Max your coding sessions are used for training while the model-improvement setting is on, and payments are generally non-refundable.

Based on Claude Code's published plans and the hands-on tests linked below, not on a test of our own.

Good for

  • Developers delegating bug fixes, tests, refactors and migrations that span many files
  • People who already pay for Claude and want one subscription for chat and coding
  • Teams automating pull request review, issue triage and recurring maintenance in CI or on a schedule
  • Non-developers ready to learn as they go who want to reach a working prototype by prompting, as Zapier recommends

Watch out for

  • Claude Code shares one usage pool with Claude chat: every paid plan has a five-hour session limit plus a weekly limit, Anthropic publishes no fixed numbers (Pro gives at least 5 times Free's per-session usage, Max 5 or 20 times Pro's), and it says it may add weekly, monthly, model or feature caps at its discretion.
  • When you hit a limit you wait for the reset, upgrade, or turn on usage credits, which bill further use at API rates from a prepaid balance up to a monthly cap you set (or no cap). An ANTHROPIC_API_KEY set on your machine makes Claude Code bill the API instead of your plan.
  • The $100 or $200 a month of API credits on Max (and Team's pooled credits) cannot be used for interactive Claude Code or for usage credits; they cover the Claude API, Managed Agents, the Agent SDK and claude -p runs with a key from the linked Console organization, and unused credits expire each billing cycle.
  • Fable models are not in Pro's or Team Standard seats' limits (usage credits only), and Max or Premium seats can spend at most half the weekly limit on them.
  • On Free, Pro and Max, Anthropic trains future models on your chats and Claude Code sessions while the model-improvement setting is on (the pricing page marks these plans opt-out) and keeps that data for five years, against 30 days with it off; Team, Enterprise and API data is not used for training by default.
  • Payments are generally non-refundable unless the Consumer Terms or local law say otherwise: you can ask support to check your eligibility, the EEA and UK get a 14-day refund prorated by use, App Store purchases are refunded by Apple, and a cancelled plan runs to the end of the period.
  • Max is billed monthly only (yearly billing is for Pro and Team, and Enterprise is yearly only), and prices vary by region and in the mobile apps.

What Claude Code does

Price

Free plan: No free plan for Claude Code: Claude's Free plan does not include it, and a guest pass from a Pro or Max subscriber gives someone new to paid Claude plans only a seven-day Pro trial that needs a card and turns into a paid plan unless cancelled.

PlanBilled monthlyBilled yearly, per monthIncludes
Pro $20 $16.67 Claude Code in the terminal, IDE, desktop and web, sharing a five-hour session limit and a weekly limit with Claude chat; Anthropic aims it at brief bursts of coding on small projects; Opus, Sonnet and Haiku included, Fable only on usage credits; $200 billed yearly
Max 5x $100 not listed 5 times Pro's usage per five-hour session plus a weekly limit, of which up to half can go on Fable; higher output limits and priority access at busy times; $100 a month of API credits for your own apps, not usable in Claude Code; monthly billing only
Max 20x $200 not listed 20 times Pro's usage per five-hour session plus a weekly limit, of which up to half can go on Fable; $200 a month of API credits for your own apps, not usable in Claude Code; monthly billing only
Team, Standard seat $25 $20 Per seat, for teams of 2 to 150: Claude Code with more usage than Pro, central billing and admin, SSO, no model training on your content by default; Fable on usage credits only; $20 a seat of pooled monthly API credits (capped at $500 a team), not usable in Claude Code
Team, Premium seat $125 $100 Per seat: 5 times a Standard seat's usage, with up to half the weekly limit on Fable; $100 a seat of pooled monthly API credits (capped at $500 a team)
Enterprise not listed $20 Per seat, billed yearly only, plus all usage billed at API rates with no per-seat usage limits; admin spend limits, SCIM, audit logs, custom data retention, a HIPAA-ready option and Code Review
Claude API (Console, pay as you go) not listed not listed Per million tokens, input/output: Opus 5.5 $4/$20, Sonnet 5.5 $2/$10, Haiku 5.5 $0.10/$0.50 (prompts up to 100K tokens), Fable 5.1 $10/$50; Anthropic puts enterprise deployments at about $13 per developer per active day, or $150 to $250 a month

Prices last checked: 2026-10-16 on Claude Code's pricing page. Full breakdown: Claude Code pricing.

Price history

No price change since we first recorded these prices on October 13, 2026. We re-check them every week and list each change here.

What hands-on testers found

We have not tested Claude Code ourselves. These are the published tests we found that used it hands-on, newest first; follow the links for the full reviews.

The New Stack · Jessica Wachtel · September 21, 2026

What they tested: Claude Code 2.1.226 on Opus 5 with her subscription against xAI's Grok Build on three memory tests in small Node repos: tell each tool a fact, quit, and give a task that depends on it in a new session.

  • Claude Code wrote each rule to a memory note right away and, in the next session, followed both in-project facts, with clearer notes than Grok's.
  • It failed the cross-project test: its memory stays inside one repo, so a rule given for all her projects did not reach a second repo unless written into ~/.claude/CLAUDE.md by hand.
  • It was faster on every session (66 seconds in all against 165) but used more tokens and cost $1.05 against $0.41, which she put down mostly to Opus 5's price; she picked Grok Build for most people.

The New Stack · Jessica Wachtel · June 8, 2026

What they tested: On a Max plan with Opus 4.8, the same goal, a codebase-health command-line tool, given twice from empty folders with a prompt written for each: once as a dynamic workflow (parallel agents) and once to a single agent, then both tools run on the HTTPie codebase.

  • Dynamic workflows split the job across five parallel agents and finished in about seven minutes with 62 passing tests; the single agent took about 11 minutes, and its session cost $2.25 by Claude Code's own count.
  • Each tool ran and flagged the same genuine weak spots in HTTPie, but the parallel version gave dependencies a perfect score because it could not read the project's setup.cfg.
  • She found the parallel result only marginally better for so small a task, said first passes still need 'an expert in the room', and concluded the feature lives up to its hype.

Zapier · Miguel Rebelo · May 7, 2026

What they tested: Claude Code and Cursor used extensively side by side on different kinds of problems, comparing setup, delegation, models, multi-agent orchestration and price.

  • Claude Code is a delegation-first agent whose harness is more mature than Cursor's for sub-agents, agent teams and CI/CD integration, with scheduled routines on top; Cursor is better for staying close to the code with tab completion and inline edits.
  • He tells non-developers to start with Claude Code and developers who want to keep writing code to pick Cursor; with the budget, use both, running Claude Code as an extension inside Cursor.
  • Once, while he kept editing the same file, he waited over 10 minutes before Claude Code said it could not change it; both tools started at $20 a month when he tested, before Cursor's October 2026 move to The X Subscription, and he calls that price increasingly an entry ticket that needs extra usage.

Tom's Guide · Amanda Caswell · February 15, 2026

What they tested: Claude Code against OpenAI's Codex on two tasks: finding three planted flaws (an SQL injection, a runaway timer and an unbounded cache) in a Node.js script, and building a sci-fi ship's terminal as a command-line app.

  • Claude Code explained its fixes clearly and was the only one to see that one buggy block should simply be deleted, but Codex added input validation that Claude Code missed.
  • Its terminal app leaned on ASCII-art headers and a flickering CRT effect; she called the creative round a toss-up.
  • She found Claude Code easier to use and better for learning, but picked Codex for building an actual application.

What users say

Test result

toolist.ai has not run its own test of Claude Code; the findings above come from the published tests linked. See how we research and test.

Best coding tools · Claude Code pricing

Do you make Claude Code? Claim this listing to correct it; it's free. To see whether ChatGPT, Claude, Gemini and Perplexity recommend it, and how many readers we send you: toolist.ai for AI companies.

Sources