Coding agent
Reads a repository and can take coding actions such as editing files, running commands or preparing a review.
Research-led directory · 85 product profiles · checked September 15, 2026
Compare terminal agents, IDE copilots, cloud software engineers, runtimes and orchestration tools. Ask in plain English, then open a sourced briefing for fit, limits and setup trade-offs.
Reads a repository and can take coding actions such as editing files, running commands or preparing a review.
Provides the loop, tools or plugins used to build and run agents; it may not be a ready-to-use coding product.
Coordinates agents, terminals or software delivery; it can support coding without being the code-writing agent.
Name check: TRAE IDE and Trae Agent are separate products. ZenCODE is a standalone CLI agent, Zen Coding extends Pi, and Zeno is a Pi desktop client. Amazon Q Developer CLI has a documented migration path to Kiro CLI, while legacy q commands remain supported. The old Windsurf Cascade docs currently redirect to Devin Desktop, so those names share one current profile. OpenCode Zen is a model service, not a coding agent ↗
1–20 of 85 matching tools · 20 per page
Terminal coding agent
Anthropic · proprietary
Anthropic’s agentic coding workflow for exploring repositories, editing code, running tools and delegating work.
Best for: Multi-file feature work and refactors
Coding agent suite
OpenAI · mixed
OpenAI’s coding-agent family spans local CLI work and delegated tasks in managed environments.
Best for: OpenAI-native teams that want local plus background coding
Agent-first development platform
Google · proprietary
Google’s terminal surface for starting, monitoring and steering Antigravity agents.
Best for: Developers already using Google’s Antigravity environment
Open-source coding agent
Anomaly · open source
A source-available coding agent with terminal, desktop and editor surfaces and broad provider support.
Best for: Switching between model providers or local inference
Minimal coding-agent harness
Pi contributors · open source
A small terminal-first harness designed to be extended with skills, themes, tools and packages.
Best for: Engineers who want a small, hackable CLI agent
Extended terminal coding agent
Community project · open source
A Pi-derived terminal agent that adds a larger toolset, LSP integration, debugging and subagent features.
Best for: Power users who want Pi-style extensibility with more built-in tools
Open-source coding agent and SDK
Cline · open source
An open-source agent available in editors, terminal and an SDK, with reviewable tool actions.
Best for: IDE-based, human-reviewed repository edits
Legacy IDE coding agent
Roo Code, Inc. · open source
A formerly popular VS Code agent with custom modes; its official GitHub repository is archived.
Best for: Understanding the evolution of mode-based IDE agents
Open-source agentic engineering platform
Kilo · open source
An open coding agent across editor and terminal surfaces with modes, model choice and agentic workflows.
Best for: Users moving between IDE and terminal workflows
Open-source coding-agent family
Qwen / Alibaba · open source
An open-source Qwen-led coding agent that supports multiple API protocols and several client surfaces.
Best for: Teams pairing open agent software with Qwen or local models
Composable agent runtime
DeepSeek · open source
A developer-preview runtime that assembles agent behavior from replaceable plugins.
Best for: Agent and harness developers exploring composable runtimes
Self-improving general agent
Nous Research · open source
A broad agent with a terminal and desktop experience, persistent skills, automations and messaging channels.
Best for: Developers who want a persistent assistant beyond code editing
Local-first AI coding and operator agent
AtomicBot · open source
An open-source, local-first agent runtime for coding and computer tasks, built around llama.cpp, a terminal UI and approval-gated tools.
Best for: Trying local GGUF models as an agent that can act on a code repository
Personal-agent gateway and orchestrator
OpenClaw community · open source
A self-hostable personal-assistant gateway that can connect messaging channels and hand work to coding agents.
Best for: A self-hosted assistant reachable through messaging channels
Multi-agent terminal workspace manager
Herdr · open source
A persistent terminal workspace that tracks and coordinates several coding-agent sessions.
Best for: Keeping multiple coding-agent terminals organized and persistent
Experimental coding-agent harness
fx.sh project · open source
A small Zig-based agent CLI and embeddable harness focused on speed, composability and local control.
Best for: Agent developers who want a tiny embeddable CLI/harness
Coding agent and private deployment platform
Poolside · mixed
A coding CLI and desktop experience backed by Poolside models and deployment options for controlled environments.
Best for: Organizations prioritizing private or controlled AI deployment
Enterprise software-delivery agent
Harness · proprietary
Governed agents for pipeline, infrastructure and release work—not a general-purpose local pair programmer.
Best for: Enterprise CI/CD, IaC and delivery operations
Open-source terminal pair programmer
Aider contributors · open source
A terminal pair-programming workflow with repository mapping, Git integration and cloud or local model support.
Best for: Terminal-first pair programming with Git as the review surface
Multi-model coding agent
Sourcegraph · proprietary
A hosted and terminal coding agent with specialist subagents and remote execution environments called Orbs.
Best for: Long-running coding tasks that benefit from remote execution
Profiles are editorial comparisons based on linked official documentation. Features, plans and availability can change; check each product’s current source before choosing.
Profiles separate verified product surfaces from editorial fit guidance. “Model approach” describes whether the product is vendor-bound, provider-flexible or a runtime; it does not promise every model will work equally well. Preview and archived entries are labeled, and official sources are linked on every profile. No benchmark rank is implied.