AI tool category
AI coding assistants: AI tools compared (31)
We checked 31 tools to write and fix code, 26 of them with a free tier. The best scored are Codex, aider, and Kilo. Median entry price: $15 per month.
What aI coding assistants tools are for
AI coding assistants help write functions, explain repositories and investigate errors. Depending on the tool, they offer completions, change files or run development commands. They can support developers, technical teams and people learning their way around an existing project. The practical change is in moving from a request to a proposal you can inspect. For example, fixing form validation previously meant locating the component, understanding its logic and drafting a change manually. With a suitable assistant, you can request a patch, examine the affected files and run relevant checks. GitHub Copilot generates functions and tests. aider edits project files through a conversation and can create Git commits. These capabilities can help prepare implementation work, but they still require technical review. Plausible code does not establish that the application behaves as intended or handles security requirements correctly.
Example uses:
- Draft a function from a specific requirement, then check its behaviour against the cases you expect.
- Understand an existing module before changing files that depend on how it works within the project.
- Investigate an error by comparing the assistant’s suggestions with the actual messages and surrounding application context.
- Generate tests for a change, then review whether they cover the risks that the modification introduces.
- Update files involved in a feature and inspect the differences before approving the proposed code change.
The main tools at a glance
Codex 90/100
Codex lets people developing software write and edit code, fix errors, run commands and tests, and review repository changes. It also supports background development tasks and programming agent integration into workflows. Features: Write and edit code files; Find and fix errors; Run commands and tests. Entry price: $8 per month, with a free tier. Runs on: web browser, Windows, macOS, Linux, iPhone and iPad, browser extension, API, and Slack. Data hosting: Several regions.
aider 78/100
aider lets developers edit project files through a conversation using local or remote models. It identifies repository structure, creates Git commits, and runs tests to fix detected errors. Features: Edit project files through a conversation; Identify the structure of a code repository; Create Git commits for changes. Free tier available. Runs on: Windows, macOS, and Linux. Data hosting: Several regions.
Kilo 78/100
Kilo supports developers writing and editing code, investigating errors, and running commands and tests. It also launches cloud agents and provides access to multiple models through one API. Features: Write and edit code; Investigate the causes of an error; Run commands and tests. Entry price: $15 per month, with a free tier. Runs on: web browser, Windows, macOS, Linux, iPhone and iPad, Android, browser extension, API, and Slack. Data hosting: United States.
GitHub Copilot 75/100
GitHub Copilot helps developers write and review code. It completes lines, generates functions and tests, answers repository questions and supports assigning changes to an agent that opens pull requests. Features: Complete lines of code; Generate functions and tests; Answer questions about a repository. Entry price: $10 per month, with a free tier. Runs on: web browser and API. Data hosting: Several regions.
Amazon Q Developer 71/100
Amazon Q Developer suggests and modifies code, explains projects and generates tests. It serves people who develop software and want help resolving AWS errors or modernizing Java applications. Features: Suggest and modify code; Explain how a project works; Help resolve AWS errors. Entry price: $19 per month, with a free tier. Runs on: web browser, Slack, and Microsoft Teams. Data hosting: Several regions.
OpenCode 71/100
OpenCode helps developers understand a project, write code, and make changes to it. It also lets them run development commands, export a session, and resume that session. Features: Explain how a project works; Write and edit code; Run development commands. Entry price: $10 per month, with a free tier. Runs on: web browser, Windows, macOS, Linux, and API.
Amp 70/100
Amp lets people developing applications modify multiple files from an instruction, run tests, and fix errors. It also supports agents on remote machines, application previews, and sharing work with a team. Features: Modify multiple files from an instruction; Run tests and fix errors; Have agents work on remote machines. Entry price: $20 per month, with a free tier. Runs on: web browser, Windows, macOS, Linux, and API. Data hosting: United States.
Continue 68/100
Continue lets people programming ask questions about project files and edit code from an editor or terminal. It runs commands, uses local or external models, and exports conversations in Markdown. Features: Ask questions about a project's files; Edit code from the editor or terminal; Run commands during a programming task. Free tier available. Runs on: Windows, macOS, Linux, and browser extension. Data hosting: On device.
Cursor 65/100
Cursor explains project code and modifies multiple files from a request. It serves people developing software who want to delegate tasks to a remote agent or propose changes on GitHub. Features: Explain a project's code; Modify multiple files based on a request; Assign a development task to a remote agent. Entry price: $8 per month, with a free tier. Runs on: Windows, macOS, Linux, web browser, and API. Data hosting: Several regions.
GoCodeo 61/100
GoCodeo helps developers create applications from instructions. The extension also lets them edit and explain code, generate and run tests, diagnose errors, and deploy projects to Vercel. Features: Create an application from an instruction; Edit and explain code; Generate and run tests. Entry price: $85 per year, with a free tier. Runs on: browser extension.
The category in figures
Of the 31 tools in the category, 26 (84%) offer a free tier, 45% have a public API and 1 have a French interface. The median entry price, per month, is $15 (across 21 tools with a published price). 18 disclose where data is hosted, 4 of them in Europe or on device.
All aI coding assistants tools
Click a header to sort. Data checked on official sites, October 24, 2026.
| Tool | Free | Entry price | API | Data | Score |
|---|---|---|---|---|---|
| Codex Codex lets people developing software write and edit code, fix errors, run commands and tests, and review repository changes. It also supports backgro | Yes | $8 per month | Yes | Several regions | 90 |
| aider aider lets developers edit project files through a conversation using local or remote models. It identifies repository structure, creates Git commits, | Yes | not published | Yes | Several regions | 78 |
| Kilo Kilo supports developers writing and editing code, investigating errors, and running commands and tests. It also launches cloud agents and provides ac | Yes | $15 per month | Yes | United States | 78 |
| GitHub Copilot GitHub Copilot helps developers write and review code. It completes lines, generates functions and tests, answers repository questions and supports as | Yes | $10 per month | Yes | Several regions | 75 |
| Amazon Q Developer Amazon Q Developer suggests and modifies code, explains projects and generates tests. It serves people who develop software and want help resolving AW | Yes | $19 per month | – | Several regions | 71 |
| OpenCode OpenCode helps developers understand a project, write code, and make changes to it. It also lets them run development commands, export a session, and | Yes | $10 per month | Yes | Not disclosed | 71 |
| Amp Amp lets people developing applications modify multiple files from an instruction, run tests, and fix errors. It also supports agents on remote machin | Yes | $20 per month | Yes | United States | 70 |
| Continue Continue lets people programming ask questions about project files and edit code from an editor or terminal. It runs commands, uses local or external | Yes | not published | – | On device | 68 |
| Cursor Cursor explains project code and modifies multiple files from a request. It serves people developing software who want to delegate tasks to a remote a | Yes | $8 per month | Yes | Several regions | 65 |
| GoCodeo GoCodeo helps developers create applications from instructions. The extension also lets them edit and explain code, generate and run tests, diagnose e | Yes | $85 per year | – | Not disclosed | 61 |
| AI2SQL AI2SQL lets people working with databases turn requests into SQL queries. It explains existing queries, corrects syntax errors, suggests optimizations | Yes | $7 per month | Yes | Not disclosed | 61 |
| Potpie Potpie helps developers ask questions about a code repository and investigate the causes of a problem. It produces tests and specifications and lets t | Yes | not published | Yes | Several regions | 61 |
| Stenography Stenography helps developers explain and document code. It adds comments in VS Code, updates documentation when saving, and suggests links to document | Yes | $10 per month | Yes | Not disclosed | 61 |
| Claude Code Claude Code supports developers in exploring and modifying code repositories. It can fix errors, edit multiple files for a task, write and run tests, | No | $200 per year | Yes | Several regions | 58 |
| JetBrains AI JetBrains AI helps developers work with code in their editor. It completes code, explains files or functions, suggests code changes, generates tests, | Yes | $100 per year | – | Not disclosed | 58 |
| Kombai Kombai helps interface developers create from instructions or turn a Figma mockup into code. It reuses existing project components and supports editin | Yes | $20 per month | – | Not disclosed | 58 |
| OpenChamber OpenChamber lets people writing code control programming agents and run multiple models on the same task. It isolates work, supports code review and a | Yes | not published | – | On device | 58 |
| Orca Orca lets people developing applications run multiple programming agents in parallel. It isolates tasks in worktrees, supports reviewing code differen | Yes | not published | – | On device | 58 |
| Devin Devin plans software changes, writes code and runs tests to correct errors. It serves people who want to work in existing repositories, prepare merge | Yes | $20 per month | Yes | Not disclosed | 56 |
| dbForge AI Assistant dbForge AI Assistant lets people working with SQL create queries from requests and adapt them to database structures. It explains SQL code, identifies | Yes | $9.95 per month | – | Not disclosed | 53 |
| Devin Desktop Devin Desktop helps developers complete and edit code in their editor. Users can assign tasks to multiple agents, find project context, review changes | Yes | $20 per month | – | United States | 53 |
| Google Antigravity Google Antigravity helps developers edit project files, run commands and execute tests. It also supports coordinating multiple development agents, che | Yes | $19.99 per month | Yes | Several regions | 53 |
| Gemini CLI Gemini CLI lets developers explore codebases and edit files from the terminal. It creates applications from instructions, runs tasks in scripts, and c | Yes | not published | – | Not disclosed | 53 |
| TRAE TRAE helps developers with code completion and questions about a project. It generates multiple files from a request, executes development tasks with | Yes | $3 per month | – | Several regions | 53 |
| Amazon Q Developer CLI Amazon Q Developer CLI lets developers chat with an assistant in the terminal on macOS and Linux. It turns requests into commands, suggests command co | Yes | $19 per month | – | Not disclosed | 53 |
| Visual Studio IntelliCode Visual Studio IntelliCode supports developers on Windows by ranking code suggestions according to context. It suggests continuations of C# lines and r | Yes | not published | – | On device | 48 |
| Cody Cody helps people developing software by answering questions about a code repository. It suggests code completions, writes and edits functions, produc | No | not published | No | Several regions | 43 |
| ZZZ Code AI ZZZ Code AI helps people writing or reviewing code. It generates code from descriptions, explains how code works, identifies errors, converts code bet | Yes | not published | – | Not disclosed | 38 |
| Gemini Code Assist Gemini Code Assist helps developers complete, generate and understand code in their editor. It suggests corrections, supports agent modifications acro | No | $0.03 pay as you go | – | Several regions | 35 |
| Qodo Qodo helps teams review code changes automatically. It identifies errors before a merge request, applies team-specific quality rules, analyzes context | No | $30 per month | – | Not disclosed | 33 |
| Metabob Metabob helps developers identify logic errors, vulnerabilities and runtime risks in code. It prioritizes issues to fix and guides coding assistants t | No | $40 per month | Yes | Not disclosed | 33 |
Also useful
These tools belong to another category but also cover this need:
- Gemini Gemini is a chatbot for people who want to ask questions, draft documents, or work on code. It also creates im
- Pieces Pieces lets people working on computers retain activity context and find information by topic or date. It summ
- Keploy Keploy supports developers by creating tests from real calls or an API specification. It simulates services an
- Claude Claude writes texts, analyzes documents and images, and writes or explains code. It serves people who want to
- v0 v0 lets people building applications start from instructions or turn a mockup into a working interface. It sup
- Hermes Agent Hermes Agent lets users execute tasks with tools and a terminal while retaining memory between sessions. It ca
- Mistral Vibe Mistral Vibe searches for information with source citations and analyzes documents and spreadsheets. It serves
- DeepCode AI DeepCode AI serves people analyzing code security within Snyk products. It identifies vulnerabilities, suggest
- Metatron Metatron helps developers extract technical decisions from a repository and its history. Proposed rules can be
- Lovable Lovable is for people who want to create a web application without coding, using written instructions. It supp
- Kimi Kimi answers questions, searches the web and analyzes long documents. It serves people who want to create pres
- HuggingChat HuggingChat is for people who want to chat with different open models or get help writing code. Users can choo
Choosing well and avoiding pitfalls
Test the assistant in your programming language, environment and a representative repository. Assess French instruction handling, change quality and how easily you can review the result. Examine context access, editor integrations and behaviour with large files. Check execution permissions, confirmation controls and ways to undo changes. Read the terms covering source code, secrets and internal information. Compare billing against users, models and usage. Prefer a workflow that preserves readable differences and supports checks suited to your project, so you can assess the proposal before it becomes part of the codebase.
An assistant can invent an API, use an outdated dependency or introduce a subtle security flaw. Check relevant documentation and review differences before approval. Generated tests can repeat the same misunderstanding as the implementation, so inspect what they actually demonstrate. Restrict command permissions and protect repository secrets. Review licences for reused material. Keep changes easy to undo, and distinguish successful execution from evidence that the application handles its intended behaviour and failure cases correctly.
To compare with the cost of using an AI model directly, see the API price calculator.
Free or paid?
Free plans often let you try completions, conversations or selected edits within usage limits. Available models, context and team features may be restricted. Paying becomes useful when the assistant has a regular role in your work or you need organisational controls. Include the time spent reviewing suggestions when assessing whether a paid plan fits your workflow.
Tools in the category with a free tier: Codex, aider, Kilo, GitHub Copilot, Amazon Q Developer, OpenCode, Amp, and Continue.
For which jobs?
The profiles that use these tools most, based on the use cases documented by publishers: developers, students, and data analysts.
Related categories
Learn more
Frequently asked questions
What is the best tool to write and fix code?
On our access and transparency grid, Codex (90/100), aider (78/100), and Kilo (78/100) come first. Test two of them on your own examples: output quality depends a lot on your case.
Are there free tools?
Yes, 26 of 31 tools have a free tier, often limited in volume: Codex, aider, Kilo, and GitHub Copilot.
Can I use a coding assistant without knowing how to program?
You can request explanations or an initial draft, but you need a way to assess the result or involve someone qualified. An application that starts can still contain logic and security problems. Begin with a bounded task and ask for an explanation of the changes before using them in practice.
Are generated tests enough to approve a code change?
Treat them as proposals that need review. Check whether they cover expected behaviour, edge cases and the risks introduced by the change. An assistant can write tests that confirm its own mistaken assumption. Combine test review with the project’s existing checks and an inspection of the code differences before approval.
Can I give an assistant access to a private repository?
That depends on your organisation’s rules and the service terms. Check what is transmitted, how it is retained and which uses are permitted. Remove secrets and restrict access to the task. If the assistant can execute commands, also inspect the accessible environment and the confirmation controls governing those actions.
Related articles
- Creating videos with AI: what works todayOctober 10, 2026
- Generating images with AI: tools, pricing, and limitationsOctober 9, 2026
- How to use NotebookLM: a complete guideOctober 8, 2026
- Free ChatGPT alternatives: what they actually offerOctober 7, 2026
Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score