AI Coding Tools

Our position. We treat the tool as interchangeable and the way of working as the part that lasts. Most of the agentic tooling a team buys in 2026 will not outlive the next model generation (opens in a new tab), while the fundamentals that help humans first and agents second will.

The form factors differ in how far the human steps back. Autonomous agents run without a human in the loop, and enterprise platforms add governance, private hosting, or codebase-scale context. The category moves faster than any other part of the AI stack. Tools that defined an earlier frontier are being restructured or retired, because model capabilities absorb what the wrapper used to provide.

Start here

• Context and productivity — How context limits tool performance and what the productivity evidence actually shows.
• Supervision and drift — How guardrails, review, skill atrophy, and the supervision paradox shape safe adoption.
• Security of autonomous coding agents — The attack surface and controls introduced by agents that can inspect and change a codebase.
• Pricing agentic coding — How subscriptions, usage units, and inference economics behave under agentic workloads.

Related

• Claude Code: one terminal agent taken apart, as the worked example of the way of working.
• AI Models and Inference: the model layer underneath every tool, and what its tokens cost.
• Productivity and Workflow: what changes in a team once the tool is installed.
• What moves adoption: why buying a tool rarely moves a team on its own.
• AI productivity studies: the measured effects behind the vendor claims.

Resources

Terminal CLIs

• Aider (opens in a new tab) - AI pair-programmer in your terminal that edits files in your git repo and commits as it goes; works with Claude, GPT, Gemini, and local models.
• Amp (opens in a new tab) - Frontier coding agent, now CLI-first after retiring its editor extensions, with automatic compaction, a plugin API, remote control, and pay-as-you-go pricing.
• Anvil (opens in a new tab) - Evidence-first coding agent built on top of GitHub Copilot CLI by the Anvil maintainer, requiring grounded references before code edits.
• Charm Crush (opens in a new tab) - Charm’s terminal coding agent with multi-model session support.
• Codebuff (opens in a new tab) - Terminal coding assistant that indexes the whole codebase and edits files or runs shell commands from natural-language instructions.
• Codebuff (Open Source) (opens in a new tab) - Open-source Codebuff with a multi-agent architecture for exploring, planning, editing, and reviewing code from the terminal.
• Command Code (opens in a new tab) - Coding agent that learns your taste via a meta neuro-symbolic model and adapts each session instead of resetting.
• Cursor CLI (opens in a new tab) - Cursor’s terminal CLI; same commands as the editor, runs in any environment.
• Gemini CLI (opens in a new tab) - Open-source agent that brings Google Gemini into your terminal.
• Groq Code CLI (opens in a new tab) - Lightweight, customizable open-source coding CLI powered by Groq for fast iteration.
• Octofriend (opens in a new tab) - Open-source coding helper CLI with multi-model support and zero telemetry.
• OpenCode AI (opens in a new tab) - Open-source coding agent for terminal, IDE, and desktop from Anomaly, with LSP autoloading, parallel sessions, 75+ model providers, and privacy-first no-code-storage design.
• Shotgun (opens in a new tab) - Spec-driven development CLI that generates codebase-aware specs and splits large features into staged, PR-sized chunks for Cursor, Claude Code, and Codex; runs locally via Ollama.
• Toad (opens in a new tab) - A universal terminal UI for agentic coding, built on the Python Textual library for flicker-free partial-region rendering.
• cplt (opens in a new tab) - Drop-in sandbox replacement for GitHub Copilot CLI on macOS.
• oh-my-pi (opens in a new tab) - Maximalist fork of pi that wires the IDE into the agent: LSP, real debuggers, persistent Python and JavaScript kernels, and content-hash edit anchors; MIT.
• pi (opens in a new tab) - Deliberately minimal open-source coding agent with four tools and a sub-1,000-token prompt, multi-provider support, and full context observability.

Editors & IDEs

• Air (opens in a new tab) - JetBrains Air, an agentic development environment where Codex, Claude Agent, Gemini CLI, and Junie run independent task loops without stepping on each other.
• Blackbox AI (opens in a new tab) - AI-native IDE and VS Code extension with multi-agent coding, real-time autocomplete across 20+ languages, image-to-code analysis, a CLI, and a unified inference API.
• Cursor (opens in a new tab) - AI-native code editor built around fast model interaction inside the editor.
• GitHub Copilot (opens in a new tab) - AI pair-programmer from GitHub and OpenAI with code completion, chat, and agent mode across VS Code, Visual Studio, JetBrains IDEs, Neovim, and the GitHub website.
• Gluecharm (opens in a new tab) - AI impact-analysis companion for spec-driven development that estimates and scopes changes across features, data models, and services before agents write code.
• JetBrains Junie (opens in a new tab) - JetBrains’ coding assistant built into the JetBrains IDEs, designed to take on tasks so you focus on results.
• Kilo Code (opens in a new tab) - Open-source coding agent for VS Code and JetBrains with access to 400+ models, specialized modes, and no vendor lock-in.
• Kiro (opens in a new tab) - AWS’s spec-driven agentic IDE, CLI, and web tool that turns prompts into EARS-notation requirements, designs, and sequenced tasks, with event-triggered agent hooks and native MCP.
• Qoder (opens in a new tab) - Alibaba’s agentic coding IDE with Agent Mode for pair programming and Quest Mode for autonomous work, supporting Claude, GPT, and Gemini.
• Tabnine (opens in a new tab) - AI code-completion and chat assistant across major IDEs, with privacy-focused self-hosted and air-gapped deployment for enterprises.
• Traycer (opens in a new tab) - Spec-first planning layer for AI coding workflows that turns ideas into step-by-step implementation plans before execution.
• Windsurf (opens in a new tab) - AI-native IDE (formerly Codeium), now part of Cognition, built around Cascade with cloud Devin agents and an Agent Command Center that manages local and cloud agent sessions in one dashboard.
• Zed (opens in a new tab) - High-performance Rust code editor from the creators of Atom and Tree-sitter, with agentic editing, parallel agents, and ACP support for Claude Agent, Codex, or OpenCode.

Autonomous Agents

• AWS DevOps Agent (opens in a new tab) - Autonomous operations teammate that investigates and prevents incidents and handles SRE tasks across AWS, multicloud, and on-premises with observability and CI/CD integrations.
• Auto Plugins (opens in a new tab) - Claude Code plugin marketplace from Auto, bundling auto-agent for workspace sync and scaffolding and ketchup guardrails that turn each AI mistake into a permanent rule.
• AutoClaude (opens in a new tab) - Experimental open-source AutoGPT-style autonomous loop for Claude via the Anthropic API.
• Base44 (opens in a new tab) - Hosted builder that turns natural-language descriptions into production-ready full-stack web apps.
• Baz (opens in a new tab) - Precision coding agents that review code, find bugs, enforce standards, and fix security issues, using vector search plus a deterministic reference graph over the codebase.
• Bolt (opens in a new tab) - In-browser AI agent from StackBlitz that generates, runs, and deploys full-stack web apps from a natural-language prompt.
• Cline (opens in a new tab) - Open-source coding agent for VS Code with Plan/Act modes, MCP integration, and terminal-first workflows.
• Cloudflare VibeSDK (opens in a new tab) - Open-source full-stack webapp generator built on Cloudflare with sandboxed previews and one-click deploy to Workers, a base for your own vibe-coding service.
• Codeligence Dev Agents (opens in a new tab) - Open-source AI teammate that hooks into Slack, GitHub, and Jira to handle docs, code reviews, debugging, and delivery automation; built on Pydantic AI.
• Devin AI (opens in a new tab) - Autonomous AI software engineer that runs parallel cloud agents on real coding tasks for engineering teams.
• Factory AI (opens in a new tab) - AI platform whose ‘Droids’ automate coding, testing, and deployment end-to-end for startups and enterprises.
• GitHub Spark (opens in a new tab) - GitHub’s prompt-to-app builder that turns natural language into full-stack apps with authentication, key-value storage, and built-in LLM prompting, deployable on GitHub’s platform.
• Google Jules (opens in a new tab) - Asynchronous coding agent from Google that runs background tasks while you focus on the work you want to do.
• Goose (opens in a new tab) - Open-source extensible AI agent from Block that installs, executes, edits, and tests with any LLM.
• Lovable (opens in a new tab) - AI app builder that turns natural-language chat into deployable full-stack web apps.
• Ona (opens in a new tab) - Background-agent platform, formerly Gitpod, with a redesigned conversation workspace, inline code review, ephemeral environments, and governance for running external agents like Claude Code.
• Open Agents (Vercel Labs) (opens in a new tab) - MIT-licensed reference app for running background coding agents on Vercel, with a web UI, durable runtime, isolated sandboxes, and GitHub App integration.
• Open SWE (opens in a new tab) - Open-source LangGraph template for internal coding agents with cloud sandboxes, Slack and Linear invocation, subagent orchestration, and automatic pull-request creation.
• OpenAI Codex (opens in a new tab) - OpenAI’s coding agent powered by ChatGPT across app, editor, and terminal, with cloud worktrees for parallel work, reusable Skills, and Automations for unprompted background tasks.
• OpenHands (opens in a new tab) - Open-source cloud coding agent platform with a model-agnostic core, sandboxed Docker/Kubernetes runtime, and native GitHub, GitLab, Slack, and CI/CD integrations.
• Replit (opens in a new tab) - Browser-based IDE and cloud platform whose Replit Agent builds, runs, and deploys full apps from natural-language prompts.
• agent-browser (opens in a new tab) - Vercel Labs browser-automation CLI for AI agents; a fast native Rust binary with accessibility-tree snapshots, batch execution, and opt-in content-boundary, domain-allowlist, and auth-vault guardrails.
• fp.dev (opens in a new tab) - Issue tracking built for coding agents, giving them scoped issues, durable context, and a CLI to keep multi-step work on track.
• mini-swe-agent (opens in a new tab) - Minimal coding agent from the SWE-agent team whose core is 100 lines with bash as its only tool, scoring over 74% on SWE-bench verified, MIT.
• openclaw (opens in a new tab) - Open-source personal AI assistant that runs on any operating system and platform.
• v0 (opens in a new tab) - Vercel’s generative-UI tool that builds React and Tailwind interfaces from natural-language prompts.

Enterprise Platforms

• AWS Q (opens in a new tab) - Amazon’s generative AI assistant that answers questions, generates content, and acts on data across AWS and enterprise systems.
• Augment Code (opens in a new tab) - Enterprise AI development tool with a context engine that indexes the full codebase for agent tasks.
• Cerebras (opens in a new tab) - Wafer-scale inference hardware delivering up to 15x faster token generation than GPUs, letting coding agents treat speed as a design parameter.
• Helix (opens in a new tab) - Open-source private agent fleet where each agent gets a GPU-accelerated desktop; runs Claude, Codex, Gemini, and open models on a self-hosted stack.
• IBM Bob (opens in a new tab) - IBM’s agentic SDLC partner with IDE and terminal integration, real-time code review, and connectors for HashiCorp, Red Hat, and Instana.
• Infercom (opens in a new tab) - EU-sovereign inference for open-source models on SambaNova dataflow hardware; GDPR-compliant, EU-hosted, claims up to 10x faster inference.
• Langdock (opens in a new tab) - German enterprise AI tool with chat, custom assistants, workflow automation, and a unified API across OpenAI, Anthropic, and Google models.
• LibreChat (opens in a new tab) - Open-source unified chat UI across multiple AI providers with agents, file handling, code interpreter, and multi-model support.
• Mistral Vibe (opens in a new tab) - Mistral’s agentic coding tool for terminal, IDE, and background agents, built on Devstral and Codestral models with fine-tuning on your code and self-hosted deployment.
• Opal (Google) (opens in a new tab) - Experimental Google Labs tool that composes prompts into multi-step mini-apps via natural language.
• SambaNova (opens in a new tab) - Custom dataflow hardware and memory architecture for fast, energy-efficient model inference and fine-tuning at data-center scale.
• Sourcebot (opens in a new tab) - Self-hosted code-understanding and search platform for humans and agents, with ask-mode over the whole codebase, MCP, and support for thousands of repos across GitHub, GitLab, and BitBucket.
• Sylph AI (opens in a new tab) - AdaL, a self-evolving coding agent for terminal and web that auto-prompts to learn your codebase and team style, with MCP tools and skills; built on the open-source AdalFlow library.
• Z.ai (opens in a new tab) - Z.ai’s GLM-4.6 chat for reasoning, coding, and agentic tasks; multimodal including vision and video.
• poolside (opens in a new tab) - Frontier lab building foundation models plus enterprise agents, TUI, and IDE surfaces deployed inside the customer’s environment for long-horizon software work.