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 vendor pages and hands-on tests published by others, not hands-on tested by toolist.ai. Updated November 3, 2026.
Best coding tools
AI coding tools range from assistants that work on your own code in an editor, through agents you hand whole tasks to, to builders that turn a prompt into a hosted app. GitHub Copilot Pro costs $10 a month. Cursor, sold since October 12 in The X Subscription, starts at $8 a month for Plus, which runs only Grok models; Premium at $30 adds cloud agents, Bugbot and other companies' models at API prices. Zed, an open-source code editor written in Rust rather than built on Electron, suits developers who want a fast native editor and their own choice of AI: the editor is free with no account, its AI runs on your own API keys, subscriptions, local models or outside agents such as Claude Agent and Codex at no charge from Zed, and Pro costs $10 a month for Zed-hosted models with $5 of tokens included. Devin Pro, which covers Cognition's Devin Desktop editor (formerly Windsurf) and its Devin Cloud agent, costs $20 a month. Claude Code comes with Claude Pro at $20 a month, and Codex with every ChatGPT plan, including Free and $8 Go, though Plus at $20 is the cheapest plan OpenAI lists with its command line, editor extension and cloud tasks. Cline, a free coding agent, open source except its JetBrains plugin, for VS Code and its forks, JetBrains IDEs, the terminal, a desktop app in beta and cloud sessions in the browser, suits developers who want to point an agent at any model: you pay your model provider, use a Claude or ChatGPT plan you already have or run a local model, and its only subscription, ClinePass, costs $9.99 a month (Cline says a processing fee may apply) for quotas on a set of mostly open-weight models. OpenCode, a free, MIT-licensed coding agent from Anomaly for the terminal, a desktop app, the web and code editors, suits developers who want a Claude Code-style agent not tied to one model company: it works with your own API keys from 75+ providers, a ChatGPT Plus or Pro or GitHub Copilot sign-in or local models, but not a Claude Pro or Max subscription. Anomaly sells models separately: Go costs $10 a month and Go Plus $40, both billed monthly and capped by a monthly dollar limit per model, and Zen is pay-as-you-go. We found no independent hands-on test of its version 2.0, which Anomaly announced on October 12, 2026, or of Go Plus. Google Antigravity is free with a weekly quota, and Google AI Pro at $19.99 a month raises its limits. Kiro, Amazon Web Services' coding agent, suits developers and teams who want an AI agent to plan before it codes: its specs turn a prompt into requirements, a design and a task list, and it works in its own IDE, a terminal CLI and, on paid plans, the browser. Kiro Free gives 50 credits a month, and Pro costs $20 a user a month for 1,000 credits, which stronger models use faster. To build an app from a prompt, Replit Core costs $20 with $20 of usage credits, Base44 Starter $20 for 100 message credits, Emergent Standard $20 for 100 credits, Figma Make comes with a Figma Full seat at $20 a month on Professional ($16 billed yearly) with 3,000 AI credits shared with Figma's other AI features, Lovable Pro and Bolt.new Pro cost $25 a month, and v0 Plus by Vercel costs $30 a user a month with $30 of usage credits.
Quick picks
- Base44: best for non-coders building internal tools and simple apps with a built-in backend
- Bolt.new: best for fast prototypes and small web apps from a chat prompt
- Claude Code: best for handing whole coding tasks to an agent from the terminal, IDE or web
- Cline: best for agentic coding on any model you choose, in VS Code, JetBrains IDEs or the terminal
- Codex: best for delegating coding tasks to an agent within a ChatGPT plan
- Cursor: best for AI agent coding in a VS Code-style editor
- Devin: best for an AI editor plus a cloud agent that opens pull requests, on one plan
- Emergent: best for non-coders building full-stack web and mobile app MVPs
- Figma Make: best for designers and product teams on Figma turning designs or prompts into working prototypes and web apps
- GitHub Copilot: best for AI coding help inside your editor and GitHub, from $10 a month
- Google Antigravity: best for Gemini coding agents across a desktop app, terminal and editor
- Kiro: best for spec-first agentic coding in an IDE, the terminal or cloud sessions
- Lovable: best for non-coders building web app prototypes and internal tools from a prompt
- OpenCode: best for model-agnostic agentic coding in the terminal, desktop app or editor
- Replit: best for building and hosting an app from a prompt without any setup
- v0: best for developers and designers building Next.js web apps and prototypes on Vercel
- Zed: best for fast native code editing with your choice of AI agents
Compared
| Tool | Cheapest paid plan | Price per unit, calculated | Basis | Prices checked |
|---|---|---|---|---|
| Base44 | Starter ($20 a month) | $0.200 per message credit | Researched | 2026-10-30 |
| Bolt.new | Pro ($25 a month) | $2.50 per million tokens | Researched | 2026-10-16 |
| Claude Code | Pro ($20 a month) | – | Researched | 2026-10-16 |
| Cline | ClinePass ($9.99 a month) | – | Researched | 2026-10-30 |
| Codex | Go ($8 a month) | – | Researched | 2026-10-16 |
| Cursor | Plus ($8 a month) | – | Researched | 2026-10-30 |
| Devin | Pro ($20 a month) | – | Researched | 2026-10-30 |
| Emergent | Standard ($20 a month) | $0.200 per credit | Researched | 2026-10-30 |
| Figma Make | Professional (Full seat) ($20 a month) | – | Researched | 2026-11-03 |
| GitHub Copilot | Pro ($10 a month) | – | Researched | 2026-10-30 |
| Google Antigravity | Google AI Pro ($19.99 a month) | – | Researched | 2026-10-16 |
| Kiro | Pro ($20 a month) | – | Researched | 2026-10-30 |
| Lovable | Pro ($25 a month) | $0.250 per credit | Researched | 2026-10-16 |
| OpenCode | Go ($10 a month) | – | Researched | 2026-10-30 |
| Replit | Core ($20 a month) | – | Researched | 2026-10-16 |
| v0 | Plus ($30 a month) | – | Researched | 2026-10-30 |
| Zed | Pro ($10 a month) | – | Researched | 2026-10-30 |
Listed in alphabetical order: we do not rank tools we have not tested ourselves. Price per unit divides each cheapest plan's monthly price by what it includes. Units differ between tools, so read the unit before comparing two rows.
Also listed, not researched yet: Griyd, Juiced, mcptask, Memdoor, QualityLayer, Reason, Skep, Smalt, Zydra.
Filter every tool by price and free plan in the tool directory, or price your own month in the AI cost calculator.
How to choose
- Editor help, coding agent or app builder: Cursor, GitHub Copilot and Zed work inside your own code, Copilot as a plugin in most popular editors, Cursor as its own VS Code-based editor and Zed as its own open-source editor, written in Rust rather than built on Electron, with a built-in agent and a panel that also runs outside agents such as Claude Agent and Codex; Claude Code and Codex are agents you hand whole tasks to from a terminal, an editor, a desktop app or the web, and Cline is an agent that runs on whatever model you connect in VS Code and its forks, JetBrains IDEs, the terminal, a desktop app in beta and cloud sessions in the browser, lets you agree a plan in Plan mode before Act mode changes anything and asks before each edit or command unless you turn on auto-approve; OpenCode, also an agent on whatever model you connect, works in a terminal, a desktop app, a web interface and extensions for VS Code, Cursor, Zed, Windsurf and VSCodium, has a Plan agent that explores and plans without editing project files, and by default lets its Build agent edit files and run shell commands without asking unless you add ask or deny rules; Google Antigravity runs several Gemini coding agents at once from a desktop app, a terminal, its own editor or extensions for other editors; Devin does both, with Devin Desktop (the Windsurf editor, renamed in June 2026) and Devin Cloud, an agent that works in its own virtual machines and opens pull requests; Kiro also does both, writing requirements, a design and a task list from your prompt before its agents code, and works in its own IDE built on Code OSS (the open-source base of VS Code), a terminal CLI and, on paid plans, Kiro Web, whose autonomous mode plans, codes and opens a pull request; Replit, Lovable, Bolt.new, Base44 and Emergent build and host a whole app from a prompt, v0 builds a Next.js web app as code you can push to GitHub and publish on Vercel, Figma Make turns a prompt or an existing Figma design into a working prototype or web app published at a figma.site address, and Emergent, like Replit and Bolt.new, also builds mobile apps, native for iOS and Android.
- Usage decides how far a plan goes: Copilot charges $0.01 a credit past its allowance, Replit bills its Agent by effort, Lovable, Base44 and Emergent charge each build message by the work it takes, Bolt.new counts tokens, v0 bills each message by the tokens it uses, including your chat history and files, Figma Make spends AI credits shared with all of Figma's AI features and Figma says it cannot know a prompt's cost until it has run (about 100 or more credits for an app from scratch with its February 2026 default model), Kiro meters everything in credits, which it defines as a unit of work, at a rate the model sets (Claude Opus 5.5 uses twice Auto's credits, GPT-5.6 Sol 4.4 times), Cursor states its plans' weekly usage only in relative terms, such as multiples of Plus (Premium 4 times, Super 15, Ultra 50), and bills Claude, GPT and Gemini separately at API list prices, Zed Pro includes $5 of tokens a month, counted at provider list price plus 10%, and past that bills usage up to a spend limit that defaults to $10 more a month, where its hosted models stop, and Claude Code, Codex, Devin and Google Antigravity run on usage limits that Anthropic, OpenAI, Cognition and Google do not state as a fixed amount of work: Claude Code on a five-hour limit plus a weekly limit, Codex on five-hour limits (currently none on Pro) and weekly limits that may also apply, Devin on daily and weekly quotas (weekly only on Max), and Antigravity on a weekly quota on the free tier, a quota refreshed every five hours up to a weekly limit on Pro and 5 or 20 times Pro's on Ultra. Past those limits, Devin's paid plans buy extra usage at API list prices and Antigravity's Pro and Ultra buy AI credits through Google One at standard API prices, while Devin Free and Antigravity's free tier wait for the reset. Kiro's monthly credits do not roll over, and past them its individual paid plans buy add-on credits at $0.04 each, twice the 2 cents a credit every paid plan works out to, while Kiro Free pauses until the next month. Cline's $9.99 ClinePass gives no quota figures beyond 2 to 5 times the usage of standard API rate limits, measured over a rolling five-hour window, the calendar week and the calendar month, and a pricier model such as Kimi K3 uses it far faster than MiMo-V2.5. OpenCode's Go ($10 a month) and Go Plus ($40), both billed monthly, give each model a monthly dollar limit counted at its token price, of which 20% can be used per five hours and 50% per week. On Go that limit runs from $15 for Kimi K3, about 110 requests per five hours by OpenCode's estimate, to $80 for DeepSeek V4.1 Flash, and on Go Plus it is 2 to 4 times Go's for the 12 paid models the Go page charts. Past its limits, Go draws on Zen credits only if you turn that on, and Zen charges per request at per-token prices OpenCode says carry no markup, plus a processing fee of 4.6% and $0.31 on each top-up. Cline, OpenCode and Zed charge nothing for AI on your own API keys, existing subscriptions (on OpenCode, ChatGPT Plus or Pro or GitHub Copilot, but not Claude Pro or Max) or local models, though with your own key your provider bills at its own token rates, so with Cline what you pay depends on the model and how many turns the agent takes. Cline says it sells its own credits at cost, and Zed Business includes no AI credit. Base44 also keeps a separate pool of integration credits for your published app's emails and AI calls, and Emergent charges credits every month to keep a published app live, from 50 a month, half of Standard's 100.
- Free plans: Copilot Free gives 2,000 code completions a month, Lovable 5 build credits a day (up to 30 a month), Base44 25 message credits a month (at most 5 a day), Figma's free Starter plan includes Make with 500 AI credits a month (at most 150 a day) shared with Figma's other AI features, publishing to the web only alongside the Figma Community, Bolt.new 1 million tokens a month, an app published on Replit's free plan goes offline after 30 days, ChatGPT Free includes limited Codex use, Devin Free gives unlimited Tab completions in Devin Desktop and a light agent quota but no Devin Cloud, Google Antigravity's free tier has every surface and unlimited Tab completions on a weekly quota, without Gemini 4 Argon, and Kiro Free gives 50 credits a month with no credit card, but no Kiro Web, cloud sessions or add-on credits; Cline itself is free, with the AI paid separately unless you run a local model or use the rotating free models a Cline account gets up to a limited quota; OpenCode itself is free and open source under the MIT License, with the AI paid separately unless you run a local model or use its rotating, limited-time free models, which need no account; Zed's free plan is the whole open-source editor with no account needed, but its AI runs only on your own keys, subscriptions, local models or outside agents, and since October 7, 2026 it no longer includes Zed's own Zeta edit predictions (other providers, such as GitHub Copilot or a local model, still work); Emergent lists 10 credits a month on its pricing page, though its help center says 10 on your first day; v0 Free allows 7 messages a day and gives new users $5 of credits once; Claude's free plan does not include Claude Code.
- Training on your code: on Copilot Free, Pro, Pro+ and Max, and on Lovable Free and Pro, you have to opt out; Cursor may use your code for training whenever Privacy Mode is off (team members have it on by default); Base44 can use your workspace data on every plan except Enterprise, with no opt-out; Figma trains on your team's files by default on Starter and Professional until an admin turns it off, and not on Organization or Enterprise; Cognition may train on Devin users' code and prompts unless a paid plan opts out (on Teams only an admin can), Devin Free cannot opt out, and Devin Enterprise data is not used without written consent; on Claude Pro and Max and on ChatGPT's Free, Go, Plus and Pro plans, Claude Code and Codex sessions can be used for training unless you turn it off, while their team, enterprise and API data is not used by default; on Antigravity's individual plans, free or paid, Google uses your sessions to improve its products and models unless you turn off telemetry; Emergent's terms let it train on your content, including your code, by default unless you opt out, and an opt-out covers only future use; v0 keeps your content out of training by default only from Business ($100 a user) up, and its plan table gives Free and Plus no data opt-out, though v0's FAQ says it honors the opt-out in a free Vercel Hobby account's data preferences; on Kiro Free and on paid plans you sign into with GitHub, Google or AWS Builder ID, Kiro may use your content for service improvement, which its docs say can include model training, unless you opt out in its settings, and its FAQ and data-protection page keep only users who sign in through IAM Identity Center or an external identity provider out of service improvement, though its enterprise page says content on Pro to Power subscriptions is not used to train foundation models; Cline's terms say it does not receive your prompts or code when you use your own key and infrastructure, but through Cline's own provider or ClinePass they go to third-party model providers that may train on them unless you opt out where they offer it, ClinePass's Muse Spark 1.3 Contributor has the name and prices of the Contributor mode of Meta's API, in which Meta's docs say your prompts and code may be used to train future Meta models, and Cline's FAQ says it never trains models on code and prompts sent through its credits, though its free-models page says free-model usage may be used to improve the models; OpenCode says it stores no code or context except when you use its free models, several of which may use what you send to improve the model, and on Go, Gemini 4 Argon Contributor and the Muse Spark Contributor models may use your prompts and outputs to train Google's or Meta's models; Zed says its contracts bar the providers of its hosted models from training on your prompts and code, it collects data to train its edit predictions only if you opt in and only from open-source projects, and with your own keys, subscriptions or outside agents the provider's own terms apply.
Read more
- Base44 review · pricing
- Bolt.new review · pricing
- Claude Code review · pricing
- Cline review · pricing
- Codex review · pricing
- Cursor review · pricing
- Devin review · pricing
- Emergent review · pricing
- Figma Make review · pricing
- GitHub Copilot review · pricing
- Google Antigravity review · pricing
- Kiro review · pricing
- Lovable review · pricing
- OpenCode review · pricing
- Replit review · pricing
- v0 review · pricing
- Zed review · pricing
Other categories: avatar video, text to video, editors, clips, script to video, voice, image, writing. All lists: best-of lists.