Pick an AI coding tool

Eleven AI coding tools, checked against each vendor's own pricing and docs pages on 2026-10-23. Answer four questions for a short list, or compare everything in the table. Names, prices and limits move fast: Windsurf became Devin Desktop in June 2026 and Gemini CLI was replaced by Antigravity, so open the source link before you pay.

Facts checked 2026-10-23. Some links on this site may become affiliate links in future; see the disclosure. None are affiliate links today.

Finder

Answer four questions, get a shortlist

Nothing leaves your browser. We rank 11 tools and show the top three with the reasons.

1. How much do you code?

2. Where do you want to work?

3. What is your monthly budget?

4. What are you building?

Answer all four questions to see your top three. Or skip ahead to the comparison table.

How the picks are ranked

Each tool gets points, and the three highest scores are shown. There is no hidden weighting and no paid placement. No link on this page is an affiliate link.

Where you work: +4 if the tool runs there, −4 if it does not (no preference scores 0). Taken from each vendor's docs.
Budget: Free only: +4 if the vendor has a free tier, −6 if not. Up to $25: +3 if a paid plan starts at $25 or less, +1 if it only has a free tier or a higher paid plan, which means you may outgrow it. Flexible: 0. Taken from each vendor's pricing page.
Experience and what you are building: each tool has a 0 to 3 score for beginners and for developers, and a 0 to 3 score for each kind of project (counted 1.5x). These scores are our own editorial judgement of fit, not benchmarks or test results.
Ties go to the lower starting price, then alphabetical order.

Prices and limits are not scored for quality. We never claim one tool is "best", only a better match for your four answers. A tool with a −4 or −6 penalty can still appear if every option has one, and we say so.

Compare all 11 tools

Every cell comes from the vendor's own pricing or docs page, checked 2026-10-23. "Not confirmed" means we could not verify it on a vendor page.

Sort by

11 tools

Tool	Runs in	Free tier	Paid from	Models	Your own API key
Bolt	Browser	Yes1M tokens/month, 300K/day, Bolt branding	$18/mo (asterisk on page)	Not confirmed which models. The page says the newest AI models are included on paid plans only.	Not confirmed
Claude Code	Terminal, Editor, Browser	NoNo free plan	$17/mo billed annually, $20 monthly	Claude models from Anthropic (Opus, Sonnet and Haiku families named on the pricing page). Docs say the terminal CLI, VS Code and JetBrains also support third-party providers.	Yes, with an Anthropic API key (ANTHROPIC_API_KEY) or a Console account
Cline	Editor, Terminal	YesFree extension; you pay for model usage (or use free models)	Free app + model usage; ClinePass $9.99/mo	Any: Anthropic, OpenAI, Google, OpenRouter, AWS Bedrock, GCP Vertex, Groq, DeepSeek and local runtimes, or Cline's own provider with free model options.	Yes
Cursor	Editor, Terminal	YesLimited agent and chat requests	$8/mo (Grok models only); $30/mo for third-party models	Cursor docs list models from Anthropic (Claude), OpenAI (GPT), Google (Gemini), Moonshot, Z.ai, Meta, plus Cursor and SpaceXAI models (Composer, Grok). Third-party models are billed at API list price.	Not confirmed
GitHub Copilot	Editor, Terminal, Browser	Yes2,000 completions/month, limited chat and agent	$10/mo	Plans page lists models from Anthropic (Claude), OpenAI (GPT), Google (Gemini), xAI (Grok), Microsoft and Kimi. Model options may vary by feature; the FAQ says Free gets entry-level models.	Not confirmed
Google Antigravity	Editor, Terminal	YesUnlimited Tab and Command requests, basic weekly agent quota	~$20/mo (Google AI Pro)	Gemini models (for example Gemini 3.8 Flash) as the core agent on all plans; third-party models are an Ultra benefit.	Not confirmed
Lovable	Browser	Yes5 daily build credits, capped at 30 a month; cannot edit or download code	$25/mo for 100 credits	Not confirmed	Not confirmed
OpenAI Codex	Terminal, Editor, Browser	YesGPT-6 Luna in the desktop app, subject to rollout	$8/mo Go; $20/mo Plus for web, CLI and IDE	OpenAI models only (GPT-6.1 Sol, GPT-6 Sol, GPT-6 Astra, GPT-6 Luna). With an API key, models follow what your key can access.	Yes, with an OpenAI API key (CLI, SDK, IDE extension)
Replit	Browser	YesStarter: daily Agent credits, Lite build, 1 published app (expires after 30 days)	$20/mo ($18 annual)	Not confirmed which models. Core and Pro include credits toward "more powerful models"; Replit AI Integrations (Core) lets your app use AI models without your own keys.	Not confirmed
v0 by Vercel	Browser	Yes7 messages a day	$30/user/mo	v0's own models (Mini, Pro, Max, Max Fast) priced per token; paid plans say access to all models.	Not confirmed
Windsurf (now Devin Desktop)	Editor, Terminal	YesLight agent quota, unlimited Tab and inline edits	$20/mo	Pro: frontier models from OpenAI, Anthropic, Google and SpaceXAI plus open-source models, and free use of SWE-2. Free has limited model availability.	Not confirmed

Each tool in detail

Strengths, limits and the pages we checked. Open a card for every plan.

Bolt

Browser

In-browser AI app builder.

Best for: Quick prototypes.

Runs in: In the browser, with website hosting and databases included.

Models: Not confirmed which models. The page says the newest AI models are included on paid plans only.

Your own API key: Not confirmed

Strengths

Quick prototypes in a browser tab
Hosting and databases included on every plan
Tokens roll over on Pro

Limits

Token use grows with project size
Free plan has a daily token cap and Bolt branding
Price carries an asterisk on the vendor page

Plans and sources

Free: $0, 1M tokens/month. 300K daily token limit, Bolt branding, public and private projects, hosting and unlimited databases.
Pro: $18*/month, 10M tokens/month with rollover. Your own domain, no Bolt branding, access to the newest models, connectors and MCPs. The page marks the price with an asterisk; see source.
Teams: $27*/month. Shared capacity and admin controls. Same asterisk on the page.

Pricing is measured in tokens.

Sources, checked 2026-10-23:

Bolt pricing

Claude Code

Terminal · Editor · Browser

Anthropic's agentic coding tool for the terminal, IDEs, desktop and web.

Best for: Developers who want a terminal-first agent that can work across a whole repo, run commands and script into CI.

Runs in: Terminal CLI, VS Code and Cursor extension, JetBrains plugin, desktop app, web (claude.ai/code), Slack, GitHub Actions and GitLab CI/CD.

Models: Claude models from Anthropic (Opus, Sonnet and Haiku families named on the pricing page). Docs say the terminal CLI, VS Code and JetBrains also support third-party providers.

Your own API key: Yes, with an Anthropic API key (ANTHROPIC_API_KEY) or a Console account

Strengths

Works across a whole codebase from the terminal
Same engine in terminal, IDE, desktop, web and CI
CLAUDE.md, skills, hooks and MCP for customizing it

Limits

No free plan
Usage limits shared with chat, five-hour and weekly windows
Terminal-first, so a steeper start for beginners

Plans and sources

Free plan: Claude Code not included. The plans comparison table lists Claude Code as No on Free.
Pro: $17/month with annual discount ($200 billed up front), $20 billed monthly. Includes Claude Code.
Max: From $100/month. 5x or 20x more usage than Pro.
Console / API: Pay as you go at API rates. Pay-as-you-go through a Console account is listed as an option for heavy use.

Claude Code is included on paid Claude plans and shares one usage pool with chat; the Free plan does not include it.
Usage limits reset on a rolling five-hour window, with weekly limits on paid plans.

Sources, checked 2026-10-23:

Cline

Editor · Terminal

Open-source coding agent for VS Code, the terminal and more; bring your own model.

Best for: Control over which model you pay for, and an open-source agent you can inspect.

Runs in: VS Code extension, CLI (npm i -g cline), desktop app; docs also list a JetBrains plugin, and Cursor, Windsurf, Zed and Neovim via ACP.

Models: Any: Anthropic, OpenAI, Google, OpenRouter, AWS Bedrock, GCP Vertex, Groq, DeepSeek and local runtimes, or Cline's own provider with free model options.

Your own API key: Yes

Strengths

Open source and model-agnostic
Every action needs your approval by default
Runs in your editor or terminal

Limits

Model usage cost is unpredictable on long tasks
You set up and manage API keys
JetBrains availability differs between the docs and the pricing page

Plans and sources

Open source: Free for individual developers; pay only for model usage. Bring your own API keys or use the Cline provider at cost. No subscription for the open-source version.
ClinePass: $9.99/month flat. Per Cline docs: 2-5x the usage on popular open coding models compared to standard API rates.
Enterprise: Contact Cline. SSO, role-based access control and governance.

Supports providers including Anthropic, Gemini, OpenAI, OpenRouter, AWS Bedrock, GCP Vertex, Groq, DeepSeek and local runtimes. GitHub reports the license as Apache-2.0.
Surfaces listed in the docs: VS Code extension, CLI, desktop app, JetBrains plugin, SDK and Kanban. The pricing page lists the JetBrains extension under Enterprise, so confirm availability.

Sources, checked 2026-10-23:

Cursor

Editor · Terminal

AI code editor and terminal agent. Plans are sold as combined Cursor and X (Grok) subscriptions.

Best for: Developers who want an editor with agentic coding built in and a choice of models from several vendors.

Runs in: Desktop editor, terminal CLI, cloud agents on web and mobile; integrates with GitHub, GitLab, Bitbucket, Azure DevOps, JetBrains, Slack and Linear.

Models: Cursor docs list models from Anthropic (Claude), OpenAI (GPT), Google (Gemini), Moonshot, Z.ai, Meta, plus Cursor and SpaceXAI models (Composer, Grok). Third-party models are billed at API list price.

Your own API key: Not confirmed

Strengths

Full editor with agent, Plan and Ask modes
Many model vendors in one place
Terminal CLI and cloud agents when you want to step away

Limits

Third-party models are billed separately from the plan
Entry plan ($8) only includes Grok models
Pricing page is a combined Cursor + X subscription, which can confuse comparisons

Plans and sources

Free: $0/month. Limited agent and chat requests.
Plus: $8/month. Shared usage pool; the page says only Grok models are available on this plan.
Premium: $30/month. 4x the usage of Plus. Adds cloud agents, AI code reviews and third-party models billed separately at API list price.
Super / Ultra: $100 and $200/month. 15x and 50x the usage of Plus.
Teams: From $30/user/month

The pricing page bundles Cursor with X and Grok features; subscriptions are sold through X Subscription Corp, not resellers.
Third-party model usage (Claude, GPT, Gemini and others) is not included in the monthly price and is billed separately at API list price.
Cursor also ships a terminal CLI (`agent`) with Agent, Plan and Ask modes, and can hand tasks to cloud agents.

Sources, checked 2026-10-23:

GitHub Copilot

Editor · Terminal · Browser

AI coding assistant inside GitHub and most major editors.

Best for: Developers who want to stay in their current editor, or who already work in GitHub issues and pull requests.

Runs in: VS Code, Visual Studio, JetBrains, Xcode, Vim/Neovim, Zed, Eclipse, Raycast and SSMS; terminal CLI; desktop app; github.com, GitHub Mobile, Slack and Teams.

Models: Plans page lists models from Anthropic (Claude), OpenAI (GPT), Google (Gemini), xAI (Grok), Microsoft and Kimi. Model options may vary by feature; the FAQ says Free gets entry-level models.

Your own API key: Not confirmed

Strengths

Works in almost any popular editor
Assign an issue to the cloud agent and get a pull request back
Lowest paid price in the group ($10/mo)

Limits

Agent and chat usage draws down AI credits that vary by model
Free plan: 2,000 completions a month and entry-level models
Individual-plan data may be used for training unless you opt out

Plans and sources

Free: $0. 2,000 completions a month and limited chat and agent usage.
Pro: $10/month. Unlimited code completions and next edit suggestions; agent mode, chat, code review and the CLI use GitHub AI Credits (1 credit = $0.01).
Pro+: $39/month
Max: $100/month

Code completions and next edit suggestions do not use credits and are unlimited on paid plans. Agent mode, code review, Copilot cloud agent, Copilot CLI and chat consume GitHub AI Credits, with usage varying by model.
Copilot Pro AI-credit totals on the plans page are $15 (Pro), $70 (Pro+) and $200 (Max) per month.
For individual plans (Free, Pro, Pro+), GitHub may use Copilot interaction data to train models unless you opt out in settings.

Sources, checked 2026-10-23:

GitHub Copilot plans

Google Antigravity

Editor · Terminal

Google's agent-first coding platform; its CLI replaced Gemini CLI for free and individual accounts on June 18, 2026.

Best for: A free way to try an agentic coding tool from Google, in the app, a terminal CLI or an IDE extension.

Runs in: Antigravity desktop app (macOS, Windows, Linux) and terminal CLI. The pricing page says IDE extensions are for Google Cloud business customers.

Models: Gemini models (for example Gemini 3.8 Flash) as the core agent on all plans; third-party models are an Ultra benefit.

Your own API key: Not confirmed

Strengths

Free tier with unlimited Tab and Command requests
Terminal CLI included at every tier
Replaces Gemini CLI, so it is where Google's free coding agent now lives

Limits

Free agent quota is a basic weekly limit and Google has changed limits before
Third-party models only on Ultra
Exact paid prices come from a Google blog post, not the pricing page text

Plans and sources

Individual: $0/month. Gemini models, unlimited Tab completions and Command requests, basic weekly rate limits.
Google AI Pro: About $20/month. Higher quota refreshed every five hours up to a weekly limit. Price from Google's Antigravity blog.
Google AI Ultra: About $100 and $200/month. 5x and 20x the Pro rate limit; third-party models are an Ultra benefit. Prices from Google's Antigravity blog.
Business: Via Google Cloud. Through Gemini Enterprise.

Gemini CLI stopped serving free and individual Google accounts on June 18, 2026; Antigravity and the Antigravity CLI replaced it, per Google and the Gemini CLI docs.
The CLI is part of every plan and uses the same quota as the app and IDE extensions.

Sources, checked 2026-10-23:

Lovable

Browser

Prompt-to-app builder with a built-in backend (Cloud).

Best for: Non-coders who want a working web app fast.

Runs in: In the browser; Git sync to GitHub, GitLab and Bitbucket on every plan.

Models: Not confirmed

Your own API key: Not confirmed

Strengths

Fast prompt-to-app for people who do not code
Git sync (GitHub, GitLab, Bitbucket) on every plan, including Free
Paid plans add code editing, code download and custom domains

Limits

Credit-based; consumption varies by prompt
Free plan cannot edit or download code and limits sharing to invited people
Models not confirmed

Plans and sources

Free: $0. 5 daily build credits, capped at 30 per month.
Pro: From $25/month for 100 credits ($250/year). Higher tiers: $50 for 200, $100 for 400, $200 for 800 credits per month.
Business: Credit tiers from 100 per month. Business plan credit rate is $0.50 per credit vs $0.25 for Pro, per Lovable docs.

Credits are used for building, chat, Cloud hosting and AI features in your apps.
Lovable's docs say you own the projects and code you build.
Free plan cannot edit or download code; editing code, downloading code and custom domains start on Pro. Git sync is on every plan.

Sources, checked 2026-10-23:

Lovable subscription plans

OpenAI Codex

Terminal · Editor · Browser

OpenAI's coding agent, included with ChatGPT plans, in the web, CLI, IDE extension and desktop app.

Best for: People who already pay for ChatGPT and want an agent in the terminal, IDE or cloud on the same plan.

Runs in: Codex web (cloud), CLI, IDE extension, desktop app and iOS on Plus and above; the desktop app only on Free and Go.

Models: OpenAI models only (GPT-6.1 Sol, GPT-6 Sol, GPT-6 Astra, GPT-6 Luna). With an API key, models follow what your key can access.

Your own API key: Yes, with an OpenAI API key (CLI, SDK, IDE extension)

Strengths

Included in a ChatGPT plan you may already have
CLI, IDE, web and cloud tasks with GitHub review via @codex
Scriptable with codex exec and the SDK

Limits

OpenAI models only
Usage is estimated, with five-hour and weekly limits
Free and Go are limited to the desktop app

Plans and sources

Free: $0/month. GPT-6 Luna at Standard speed in the desktop app, subject to rollout.
Go: $8/month. Lightweight coding tasks; desktop app, subject to rollout.
Plus: $20/month. Codex on the web, CLI, IDE extension and iOS, with GPT-6.1 Sol and GPT-6 Luna.
Pro: $100, $200 or $500/month. Choose the Pro plan that fits your usage.
API key: Pay at API pricing. CLI, SDK and IDE extension only; no cloud features.

ChatGPT Work and Codex share one usage allowance.
Plus: OpenAI estimates 15-160 local messages per five hours on GPT-6.1 Sol and 350-3,000 on GPT-6 Luna; weekly limits may also apply. Pro plans currently have no five-hour limit.
API-key use has no cloud features (GitHub code review, Slack) and is billed at API pricing; GPT-5.5 retired from Codex with ChatGPT sign-in on October 14, 2026.

Sources, checked 2026-10-23:

Codex pricing

Replit

Browser

Browser-based builder with an Agent and built-in hosting.

Best for: Going from a prompt to a deployed app without leaving the browser.

Runs in: In the browser, with built-in hosting, databases and publishing.

Models: Not confirmed which models. Core and Pro include credits toward "more powerful models"; Replit AI Integrations (Core) lets your app use AI models without your own keys.

Your own API key: Not confirmed

Strengths

Prompt to deployed app in one browser tab
Hosting and databases built in
No local setup

Limits

Starter has Lite build only and the free published app goes down after 30 days
Credit-based usage can exceed the plan price
Exact models not confirmed

Plans and sources

Starter: Free. Daily Agent credits up to a monthly cap, Lite build only, monthly cloud credits, 1 published app that goes down after 30 days.
Core: $20/month ($18/month billed annually). Up to 30 hours of chat on Free Mode, $20 toward more powerful models, unlimited workspaces.
Pro: $100/month ($90/month billed annually). Up to 15 collaborators, $100 toward the most powerful models, 28-day database rollback.

Starter is limited to Lite build; Full build, Plan Mode, third-party connectors and removing the "Made with Replit" badge need Core.
Core adds pay-as-you-go usage on top of its included credits; databases, storage and publishing draw on monthly credits.

Sources, checked 2026-10-23:

v0 by Vercel

Browser

AI UI and app generator from Vercel.

Best for: Generating React/Next.js UI you then own.

Runs in: In the browser at v0.app; deploys to Vercel and syncs with GitHub.

Models: v0's own models (Mini, Pro, Max, Max Fast) priced per token; paid plans say access to all models.

Your own API key: Not confirmed

Strengths

Deploys to Vercel and syncs with GitHub
Design Mode for visual edits
Its own models (Mini, Pro, Max) with published per-token prices

Limits

Free plan is 7 messages a day
Cheapest paid plan is $30 per user
Usage is billed as credits per token, so long prompts cost more

Plans and sources

Free: $0/month
Plus: $30/user/month. Includes $30 of monthly credits and $2 of free daily credits per user.
Business: $100/user/month

Credits can be bought on top and are shared across the team.
Free plan includes deploying to Vercel, GitHub sync and Design Mode, with a 7 messages per day limit.

Sources, checked 2026-10-23:

v0 pricing

Windsurf (now Devin Desktop)

Editor · Terminal

AI code editor. Renamed Devin Desktop by Cognition on June 2, 2026; plans and pricing carried over.

Best for: Developers who want a full VS Code-style editor with an agent manager and a $20 paid tier.

Runs in: Desktop IDE (VS Code compatible) with an Agent Command Center; a separate Devin CLI and Devin Cloud agent exist in the same product line.

Models: Pro: frontier models from OpenAI, Anthropic, Google and SpaceXAI plus open-source models, and free use of SWE-2. Free has limited model availability.

Your own API key: Not confirmed

Strengths

Full editor plus a view for managing several agents
Unlimited Tab completions on the free plan
Works with local-only agents

Limits

Rebranded June 2026, so older guides use the Windsurf name
Jump from $20 Pro to $200 Max
Free plan has a light agent quota and limited models

Plans and sources

Free: $0. Light quota for agents, limited model availability, unlimited Tab completions and inline edits.
Pro: $20/month. Higher quotas with frontier models from OpenAI, Anthropic, Google and SpaceXAI plus open-source models; free use of SWE-2; extra usage at API pricing.
Max: $200/month. Significantly higher quotas than Pro.
Teams: $80/month for the team plan + $40/month per full dev seat

Cognition renamed Windsurf to Devin Desktop on June 2, 2026; the app, plans and pricing carried over, and windsurf.com still shows the pricing.
The product line now includes Devin Desktop (editor with an Agent Command Center), Devin Cloud, Devin CLI and Devin Review.

Sources, checked 2026-10-23:

Which one fits you?

Never coded and want an app online this week? Start with a browser builder (Replit, Lovable, Bolt or v0). Nothing to install, but they run on credits or tokens, and you may need a paid plan to edit or download the code.
Already code and live in an editor? GitHub Copilot works in a long list of editors and starts at $10 a month. Cursor and Devin Desktop (formerly Windsurf) are editors built around an agent.
Prefer the terminal, scripts or CI? Claude Code, OpenAI Codex, Cline and the Cursor and Antigravity CLIs all run there.
Want to pick (and pay for) the model yourself? Cline is open source and works with your own keys, so you pay the model provider directly.
Already pay for ChatGPT, Claude or a Google AI plan? Check first: Codex is included with ChatGPT plans, Claude Code with paid Claude plans, and Antigravity's paid tiers are Google AI plans.
Tight budget? Ten of the eleven have some kind of free tier. Claude Code is the exception, with no free plan. Free tiers are small, so treat them as a trial.

Our picks

You already work in VS Code, JetBrains or another editor

GitHub Copilot

Why: Works in a long list of editors, plus a CLI and GitHub itself, with a free plan and a $10 a month paid plan.

Tradeoff: Agent and chat usage draws down AI credits, and how fast depends on the model you pick.

Source: github.com · checked 2026-10-23 · full entry

You like the terminal and want an agent that works across a whole repo

Claude Code

Why: Runs in the terminal, VS Code, JetBrains, the desktop app and the web on one engine, and can run in CI.

Tradeoff: No free plan, and usage is shared with Claude chat in five-hour and weekly windows.

Source: claude.com · checked 2026-10-23 · full entry

You want an AI-first editor with models from several vendors

Cursor

Why: A full editor with agent, plan and ask modes, a terminal CLI and cloud agents, with models from Anthropic, OpenAI, Google and others.

Tradeoff: Third-party models are billed separately from the plan, and the $8 Plus plan only includes Grok models.

Source: cursor.com · checked 2026-10-23 · full entry

You want to choose the model and see every step

Cline

Why: Open source, works with your own API keys or local models, and asks for approval before actions.

Tradeoff: Model costs are usage-based and can climb on long tasks, and you manage the keys.

Source: cline.bot · checked 2026-10-23 · full entry

You already pay for ChatGPT

OpenAI Codex

Why: Included in ChatGPT plans, with a CLI, IDE extension and cloud tasks on Plus ($20 a month) and above.

Tradeoff: OpenAI models only, and included usage is an estimate that depends on model and task.

Source: developers.openai.com · checked 2026-10-23 · full entry

You want a free Google option in the terminal

Google Antigravity

Why: Free individual tier with unlimited Tab and Command requests, and the CLI is included at every tier.

Tradeoff: The free agent quota is a basic weekly limit, and Google has changed these limits before.

Source: antigravity.google · checked 2026-10-23 · full entry

You have never coded and want a live app

Replit

Why: Prompt to published app in the browser, with hosting and databases built in and a free Starter plan.

Tradeoff: Starter is Lite build only, and its free published app goes down after 30 days.

Source: replit.com · checked 2026-10-23 · full entry

Common mistakes

Comparing plan prices and ignoring model costs

Cursor bills third-party models separately from the plan, Copilot spends AI credits at a rate that depends on the model, and Cline passes through whatever your model provider charges. Compare the all-in monthly cost for how you will actually work.

Source: cursor.com/pricing

Trusting a 2025 review

Windsurf was renamed Devin Desktop on June 2, 2026, and Gemini CLI stopped serving free and individual Google accounts on June 18, 2026, when Antigravity replaced it. Plans and model lineups have moved too, so check the date on any comparison.

Source: devin.ai/blog/windsurf-is-now-devin-desktop

Assuming free means unlimited

Lovable Free gives 5 build credits a day, capped at 30 a month. v0 Free is 7 messages a day. Bolt Free has a 300K daily token limit. Replit Starter publishes one app that goes down after 30 days. Use free tiers to test, not to run a project.

Source: docs.lovable.dev/introduction/subscription-plans

Using a browser builder when you need to own the code

On Lovable Free you cannot edit or download code. On Replit Starter, Agent is limited to Lite build. If you plan to hand the project to a developer later, check code export and Git sync before you start.

Source: docs.replit.com/billing/plans

Paying twice for access you already have

Codex is included in ChatGPT plans, Claude Code is included in paid Claude plans, and Antigravity's paid tiers are Google AI plans. Look at what your current subscriptions already cover before adding another.

Source: developers.openai.com/codex/pricing

Letting an agent run commands unreviewed

Cline asks for your approval before every action by default. Whatever tool you use, keep that kind of review on until you trust the setup, and keep secrets out of prompts.

Source: docs.cline.bot/getting-started/what-is-cline

All sources for this page