159 tools
Agents & MCP
Coding agents, assistants, MCP servers, and agent skill directories in one stack.
github.com icon
OpenAI's local terminal coding agent (Rust) that reads repos, proposes patches, and runs shell commands; pairs with
windsurf.com icon
Cognition's agentic IDE (ex-Codeium Windsurf) with Cascade-style flows and in-editor autonomous agents; marketed
Microsoft/GitHub's coding assistant across IDE plugins and Copilot CLI/agent modes—completions, chat, and multi-step
GitHub's issue-to-PR and agentic workspace flows that plan, edit, and open pull requests from natural-language tasks
devin.ai icon
Cognition's cloud autonomous software engineer that plans, codes, debugs, and deploys in a sandboxed environment with
aider.chat icon
Open-source git-native CLI pair programmer with repo maps, automatic commits, and multi-model support for surgical
continue.dev icon
Open-source VS Code & JetBrains assistant for chat, autocomplete, edit, and agent modes with full model/provider
cline.bot icon
Open-source autonomous coding agent as VS Code/JetBrains extension, CLI, and SDK—file edits, terminal, browser tools
roocode.com icon
Popular Cline fork / VS Code agent with specialized modes (code, architect, debug) and multi-mode autonomous workflows
opencode.ai icon
MIT open-source terminal coding agent (TUI/desktop/IDE) that is model-agnostic across 75+ providers including local
Block-originated open-source local-first agent (desktop + CLI + API) for code and general workflows with deep MCP
ampcode.com icon
Sourcegraph's opinionated agentic coding product (VS Code extension + CLI) emphasizing frontier models, shareable
Showing 1–12 of 159
1
About this shelf
Coding agents and AI IDEs are the tools that edit a real repo: files, tests, terminals, and instruction files like AGENTS.md. This shelf also absorbs assistants, MCP servers, and skill directories so the agent stack lives on one page. If a product only pastes snippets into chat, it does not belong here.
Pick by loop. Some people want an AI-native IDE all day. Some want a terminal agent over the whole tree. Some want autocomplete with corporate approval. Mixing Cursor for daily work and a CLI agent for refactors is normal. The compare guides for Cursor vs Claude Code and Copilot vs Cursor exist so this page does not have to be a holy war.
Filters split assistants, MCP, and skills. Use them when you already know you need a server or a SKILL.md directory. Featured picks are the daily drivers we would actually pay for. The grid is the long tail, including smaller MCP and skill listings that never appear in IDE reviews.
Guardrails matter more than the model name this week. Tests, types, and a short instruction file turn these from slot machines into teammates. Read the glossary entries for AGENTS.md, CLAUDE.md, SKILL.md, and MCP if the words on the cards are still mush.
Questions
- What counts as an agent here?
- A product that can touch a codebase or expose tools to one — IDEs, terminal agents, MCP servers, and skill directories we list.
- Where are MCP servers?
- On this shelf under the MCP filter, plus a glossary page that explains the protocol.
- Should I pick Cursor or Claude Code?
- Read /compare/cursor-vs-claude-code. Many teams use both: IDE for daily edits, CLI for long refactors.
- Do you noindex tool pages?
- No. Each listed agent has an indexable /tool page with alternatives and catalog fields.