AI Coding Agent on Vibe Coding

Replit Agent

Mon, 01 Jan 0001 00:00:00 +0000

Replit Agent works inside the browser: it can create a project from scratch, write the code, install packages, provision a database, and deploy — with a live preview updating as it works.

Bolt.new

Mon, 01 Jan 0001 00:00:00 +0000

Bolt.new runs entirely in the browser using WebContainers, so the generated app boots, installs npm packages, and runs without any server round-trip. You can prompt changes and watch the live app update instantly.

Lovable

Mon, 01 Jan 0001 00:00:00 +0000

Lovable is built for people who want to go from an idea to a working product without writing code themselves — you describe the app conversationally, and it scaffolds the frontend, backend, and database, with a visual editor layered on top for tweaks.

Devin

Mon, 01 Jan 0001 00:00:00 +0000

Devin, built by Cognition, is positioned as an autonomous software engineer rather than an in-editor assistant: hand it a ticket-sized task and it plans the work, writes and tests code in its own cloud environment, and comes back with a pull request rather than requiring turn-by-turn direction.

Amazon Q Developer

Mon, 01 Jan 0001 00:00:00 +0000

Amazon Q Developer is AWS’s coding assistant, combining inline code completion with an agentic mode that can plan and execute multi-file changes, review code for issues, and automate tasks like dependency upgrades — all with native awareness of AWS services and infrastructure.

Factory (Droid)

Mon, 01 Jan 0001 00:00:00 +0000

Factory positions its agents — “droids” — as configurable workers for an engineering organization rather than a single personal assistant: separate droids can be set up for coding, code review, incident response, or release management, each usable from the terminal or CI.

OpenHands

Mon, 01 Jan 0001 00:00:00 +0000

OpenHands (formerly OpenDevin) is an open-source platform for running autonomous coding agents that execute real engineering work — planning a task, writing code, running commands, and applying changes — rather than just suggesting edits, with a composable SDK for building custom agent workflows on top of it.

Augment Code

Mon, 01 Jan 0001 00:00:00 +0000

Augment Code targets a specific problem: giving an agent reliable context across codebases too large and interdependent for most tools to index well. It combines an editor extension with an agent that can plan and execute multi-file changes grounded in that broader context.

Cosine

Mon, 01 Jan 0001 00:00:00 +0000

Cosine builds an AI software engineer, Genie, that plugs into a team’s existing repos and issue tracker to pick up tickets, work through a codebase, and submit pull requests for review rather than just suggesting snippets inline.

SWE-agent

Mon, 01 Jan 0001 00:00:00 +0000

SWE-agent wraps an LLM in an “Agent-Computer Interface” — a constrained set of commands for viewing, searching, and editing files and running tests — so it can work through real GitHub issues end-to-end. It’s the reference agent behind much of the SWE-bench benchmark work.

Softgen

Mon, 01 Jan 0001 00:00:00 +0000

Softgen generates full-stack web apps — UI, backend logic, and database — from a chat interface, aimed at founders and builders who want to go from idea to working product without hand-writing the initial implementation.

Create

Mon, 01 Jan 0001 00:00:00 +0000

Create turns a written description of an app into a functioning build — pages, interactions, and backend data — that updates as you keep describing changes, without requiring the user to touch code.

Base44

Mon, 01 Jan 0001 00:00:00 +0000

Base44 lets non-developers describe a product idea in chat and get back a working, hosted application with a database, authentication, and integrations already wired up, without touching code or assembling separate services.

HeyBoss

Mon, 01 Jan 0001 00:00:00 +0000

HeyBoss positions itself as an AI “product team in a chat window” — describe a website, app, or landing page and it generates, hosts, and lets you keep iterating on it conversationally, aimed squarely at non-developers.

Emergent

Mon, 01 Jan 0001 00:00:00 +0000

Emergent builds working full-stack applications from a described idea, handling frontend, backend, and data layer generation in one flow so builders can go from prompt to a functioning app without assembling separate tools.

Databutton

Mon, 01 Jan 0001 00:00:00 +0000

Databutton pairs a chat-driven agent with a workspace for building internal tools, dashboards, and small full-stack apps, generating Python backends and React frontends and handling hosting so teams can ship without setting up infrastructure.

Val Town

Mon, 01 Jan 0001 00:00:00 +0000

Val Town lets you write a small TypeScript script in the browser and run it instantly as a hosted endpoint, cron job, or webhook — no deploy pipeline required — with an AI assistant that can generate and modify vals from a prompt.

Magic

Mon, 01 Jan 0001 00:00:00 +0000

Magic is developing its own frontier models trained specifically for software engineering, with a focus on extremely long context windows so an agent can reason over an entire large codebase at once. It’s earlier-stage and less publicly available than most tools in this list, but notable for the scale of investment behind an autonomous-engineer approach.

Devika

Mon, 01 Jan 0001 00:00:00 +0000

Devika emerged as an open-source answer to Devin — an agent that takes a high-level coding objective, plans the steps to get there, browses for information it needs, and writes and iterates on code with visibility into its reasoning at each step.

MetaGPT

Mon, 01 Jan 0001 00:00:00 +0000

MetaGPT assigns different LLM agents distinct roles — product manager, architect, project manager, engineer — and has them collaborate through standard software-team artifacts like PRDs and design docs on the way to generating a working codebase from a single requirement.

bolt.diy

Mon, 01 Jan 0001 00:00:00 +0000

bolt.diy takes StackBlitz’s Bolt.new — an AI app builder that runs a full dev environment in the browser via WebContainers — and opens it up so anyone can plug in their own model, including local or open-weight ones, instead of being locked to a single provider.