AI agents for software development
Coding agents write, review and change code, work on issues and operate in the terminal, editor or repository – from assistants to largely autonomous development agents.
Software development is the field in which AI agents are most widespread so far. Coding agents read a repository, plan changes, run commands in an isolated environment and propose pull requests.
The range spans assistants in the editor, agents in the terminal and systems that pick up tasks from an issue tracker and run several agents in parallel.
Role in an AI-first business model
For AI-first companies, coding agents mainly change the speed and cost of their own development. Processes, integrations and internal tools can be adapted faster – provided that review, testing and accountability for the code remain clearly defined.
Software development also illustrates what an AI-first operating model is about: people set the goal and the quality standard, agents carry out the work, and defined checkpoints decide what is accepted.
What to look for when choosing
- Isolated execution (sandbox) for commands and tests
- Traceable changes via pull requests and review
- Clear permissions: what may the agent change without approval?
- Model choice and cost control for large codebases
56 tools in this category
MetaGPT
Multi-agent framework in which several AI agents work together in roles modelled on a software company. The aim is to produce software from requirements written in natural language.
OpenHands
Open-source platform for AI-powered software development agents, formerly known as OpenDevin. Useful for letting coding agents handle development tasks with a high degree of autonomy.
GPT Pilot
Core technology behind the Pythagora extension for VS Code. The project aims to provide an AI developer that works alongside developers while they build applications.
Aider
Tool for AI pair programming directly in the terminal. Developers work on their code in dialogue with a language model and have it make changes in the local repository.
Devika
Agentic AI software engineer that breaks high-level instructions into individual steps, researches the relevant information and then writes code to reach the stated objective.
RepoAgent
LLM-powered agent for code repositories. It helps developers and teams generate documentation and get up to speed with existing repositories quickly.
DSPy
Framework from the Stanford NLP group for programming foundation models rather than prompting them by hand. Useful for structured, repeatable LLM applications and pipelines.
ThinkGPT
Python library from Jina AI with agent techniques intended to extend what a language model can do. It is aimed at developers who want to use LLMs beyond simple prompts.
PyCodeAGI
Small experiment that generates a Python app from a description of the application the user wants. Better suited to trying things out and learning than to production use.
SuperAGI
Open-source framework for autonomous AI agents, explicitly aimed at developers. It is used to build, manage and run your own agents.
Plandex
AI coding engine for larger, complex development tasks. It is suited to situations where changes need to be planned and carried out across many steps and files.
Codel
Autonomous AI agent that uses a terminal, browser and editor to work on complex tasks and projects. This allows multi-step development work to be handled largely on its own.
DB GPT
Project for working with your own data and environment through locally run language models. Its focus is on keeping data within your own infrastructure.
Agency
Go library for developers who want to bring large language models and other generative AI into their applications. It follows a lean, idiomatic Go programming style.
MicroAgent
Experimental project with agents that can edit their own prompts and Python code. Of interest for exploring agents that improve themselves.
SWE Agent
Coding agent that takes a GitHub issue and attempts to fix it automatically. The underlying language model can be chosen freely.
AgentRun
Tool for running AI-generated Python code in an isolated, safe way. Useful when agents produce code that should not run unchecked on your own system.
Claude Engineer
Interactive command-line application that uses an Anthropic Claude model for software development tasks. Developers work with it in dialogue directly in the CLI.
Vision agent
Library from LandingAI that uses agent frameworks to generate code for computer vision tasks. Useful for implementing vision solutions more quickly.
TypedAI
TypeScript platform for AI agents, including autonomous agents, software development agents and agents for code review.
Cline
Open-source coding agent that gives developers direct access to current frontier models. It keeps what the agent does transparent.
OpenCode
AI coding agent built specifically for the terminal. Suited to developers who want to hand coding tasks to an agent directly from the command line.
Stakpak
Open-source DevOps agent that helps secure, deploy and maintain production infrastructure.
ReviewCerberus
Free, open-source tool for AI-assisted code review. It analyses the differences between Git branches for security, performance and code quality.
Dorothy
Open-source desktop app that runs several command-line AI coding agents such as Claude Code, Codex or Gemini in parallel. It includes automations, a Kanban board, remote control and MCP servers.
Greywall
Deny-by-default command sandbox for AI coding agents. It isolates the file system, controls network access and ships with profiles for agents such as Claude Code or OpenCode.
Maestro Orchestrate
Platform for orchestrating specialised AI agents in multi-phase development workflows with parallel execution and tiered permissions. It runs on Gemini CLI, Claude Code and Codex.
amux
Agent multiplexer built on Python and tmux for running many Claude Code sessions in parallel. It provides a web dashboard, a Kanban board and a REST API for agent-to-agent communication.
AgentsMesh
Platform for remote AI workstations ("AgentPods") with a PTY sandbox and git worktree isolation. Agents collaborate through channels, and a built-in Kanban board with MR/PR integration keeps track of the work.
KubeStellar Console
Multi-cluster Kubernetes dashboard with an AI-powered operations agent (kc-agent). Its MCP tools help manage workloads across edge and cloud clusters; the project is part of the CNCF.
Bernstein
Open-source governance layer for AI agents. It suits teams that want to put clear rules and controls around how their agents operate.
ctop
An htop-style monitor for coding agents. It shows real-time CPU, memory, token usage, context window and estimated cost for Claude Code and Codex CLI sessions, written in pure Node.js without dependencies.
Frontman
A coding agent that runs in the browser and connects to your dev server. It sees the live DOM, component tree, CSS, routes and logs, and edits the actual source files with instant hot reload.
Maestro
Desktop command center for running several coding agents such as Claude Code, Codex or Gemini CLI in parallel. It adds event-driven automation, run playbooks, group chat across local and remote agents, and a CLI the agents can drive themselves.
Cate
Desktop IDE on an infinite, zoomable canvas. Editors, terminals, browsers and Claude Code panels are arranged spatially instead of in tabs, can dock or detach into separate windows, and the layout is saved per project.
Nanocoder
Local-first command-line coding agent built with React and Ink. It is aimed at developers who want a coding agent in the terminal on their own machine.
zeroshot
CLI that coordinates a planner, an implementer and independent validators in isolated environments. This keeps planning, implementation and validation of code changes clearly separated.
h5i
CLI that has several coding agents tackle the same task, each in its own git worktree. The agents review each other, and a neutral verifier runs the tests and merges the candidate that passes.
Paperclip
Open-source platform for running coding agents as managed workers. Tasks, approvals, budgets and workspaces bring structure and control to how agents are used.
Agent Teams AI
Open-source desktop orchestrator for autonomous teams of coding agents. It offers task delegation, messaging between agents, Kanban and code review across multiple agent runtimes.
Tura
Local open-source coding agent designed to understand a repository before changing it. It is aimed at developers who want traceable agent behavior rather than unsupported claims.
Orca
Terminal coding agent written in Rust and built for DeepSeek models. It includes OS-level sandboxing, a persistent goal mode, background tasks and verification gates.
fractal
Framework for hierarchical agent loops that organize themselves recursively. Useful for experimenting with nested agent structures.
OpenCodeReview
Open-source code reviewer that combines deterministic rules with an LLM agent that uses tools and leaves line-level comments. It can be self-hosted and run with a model of your choice.
OpenBitFun
Open-source desktop agent with a Rust runtime that works across real repositories, the browser, the terminal and the desktop. It can be extended through MCP, Skills and custom agents.
CompozyOS
Open-source "operating system" for AI agents: plug in the agent CLIs you already use and run them as a team on loops and schedules. Shared memory, permissions and approvals are built in.
Ouroboros (Agent OS)
Agent system that first refines the spec through a Socratic interview and only approves it once ambiguity is low enough. A single command then drives execution, staged evaluation and a budgeted improvement loop.
Atomic Agent
Local-first coding agent with a CLI and TUI that runs open-weight models entirely on your own machine. It is based on a fork of llama.cpp.
Keen Code
Context-aware open-source terminal coding agent written in Go. It supports multiple providers, controllable retention of tool output across turns, skill-driven MCP integration, subagents and Agent Skills.
Ordewell
Terminal CLI and TUI that turns a goal into an ordered, editable plan of coding-agent tasks. Each task runs as its own session on the harness and model you pick, and counts as done only when its unique completion marker appears.
Orbi
Self-hosted autonomous coding agent that takes a labelled GitHub issue all the way to release: implementation, an independent review session, merging the exact reviewed head, and a tagged release.
agent-manager
Terminal UI that runs coding-agent CLIs side by side, each in its own persistent tmux session. Live status, quick prompts, git worktrees and diff review make parallel work easier.
mu
Coding agent built on Pi with a CLI and desktop app, where a small judge model makes routine calls, such as which tool output enters the context or whether a result carries instructions aimed at the model. Every verdict is logged locally.
Claude Code Local
Project that runs the Claude Code agent fully on-device on Apple Silicon. A local MLX server speaks the Anthropic Messages API, and a parser handles the tool-call formats of local models.
wmux
Desktop workspace for Windows and macOS that runs Claude Code, Codex, Gemini CLI and other CLI agents side by side, with up to eight parallel tasks in their own git worktrees. It adds hunk-level diff adoption, approval gates for tool calls and sessions that survive reboots.
ReevesAgents
Local workspace that runs coding-agent CLIs side by side in tmux, controlled from a TUI, a local web UI or the CLI. An opt-in MCP server lets one agent spawn, read and steer the others.
From tool to outcome
In AI-first coaching you clarify which outcomes AI agents should be accountable for in your go-to-market and within which framework. REVAGO then designs, builds and runs your system. You pay for results – not for activity.
Key concepts
Other categories
A neutral overview, not a recommendation of individual providers. REVAGO is not affiliated with the listed projects. Descriptions are based on the projects’ public information; features and licences may change. Selection based on awesome-agents (kyrolabs), As of 19 October 2026.