Cloud Parts

Terminal agents

Agents you start in a shell, in your repository, that use the tools already installed on your machine.

How they work

A terminal agent runs where you run git and your test suite. It reads files in the current directory, runs commands you could run, and shows its tool calls and diffs as it goes. Claude Code, OpenAI’s Codex CLI, Gemini CLI and OpenCode all work this way, and most now also ship editor extensions or desktop apps that drive the same loop.

Because they’re command-line programs, they also run without a person: claude -p, codex exec, gemini -p and opencode run take one prompt, work, and exit, so they fit into scripts and CI jobs. The trade-off is that they act on your real machine, so permission rules and an OS sandbox matter more than for a cloud agent with its own VM. When you close the laptop, the work stops.

Technologies

What all ai coding agents have in common →

Terminal agents

Claude CodeCommercialAnthropic’s coding agent: a terminal program, also available in IDEs, a desktop app and the cloud, that drives Claude models through a loop of reading, running and editing.Gemini CLICommercialGoogle’s open-source terminal agent for Gemini models; since 18 June 2026 it serves only paid API keys and business licences, as Google moved free and consumer users to Antigravity CLI.OpenAI CodexCommercialOpenAI’s coding agent: an open-source CLI, written in Rust, that works in your local repository inside an OS sandbox, plus an IDE extension and Codex cloud for tasks that run on OpenAI’s machines.OpenCodeOpen sourceAn open-source (MIT) coding agent that works with 75+ model providers and local models, built as a local server with a terminal interface as one of its clients.+1 Kilo Code

Worth exploring

All resources →
Video· August 2025Advanced Context Engineering for AgentsDex Horthy (HumanLayer), YC Root AccessA talk on getting coding agents to work in large, existing codebases by keeping their context small and correct: split work into research, plan and implement, compact each step into a Markdown file, keep context use around 40–60%, and spend human review on the research and plan rather than every line of code. Summarised here from the written version in HumanLayer’s GitHub repository, which also notes where the approach failed; HumanLayer builds tools for this workflow.

Sources

Improve this page on GitHub