Choosing an AI Coding Agent

The first version of this comparison tried to describe every product with a list of strengths, traps, caveats, and biases. The result was thorough and not very useful. The practical question is smaller: which tool should I try first for the way I work, and what would make me switch?

What the public numbers say

Terminal-Bench 2.0 tests multi-step work in isolated terminal environments. These were representative entries visible in the 2026-08-10 snapshot:

Harness	Model	Harness version	Accuracy	Verified
Codex CLI	GPT-5.5	0.121.0	82.2%	yes
Gemini CLI	Gemini 3.1 Pro	0.35.0	61.4%	no
Claude Code	Claude Opus 4.6	2.1.34	58.0%	yes
OpenCode	Claude Opus 4.5	not stated	51.7%	no

Codex had the strongest result among these named configurations. That makes it a sensible first candidate for terminal-heavy work. It does not prove that the Codex harness alone caused the difference: the model, version, prompt, permissions, and date all changed across rows.

Cursor and Pi did not have entries in that snapshot. Cursor is centered on an IDE workflow that a terminal benchmark barely measures. Pi can be extended into many different harnesses, so “Pi” without its model and extensions is not one stable test subject.

That is the only benchmark warning worth repeating. The table narrows a shortlist; it does not finish the choice.

What each tool buys

Tool	Why I would choose it	What I would watch
Claude Code	cohesive repository workflow with instructions, hooks, MCP, permissions, and subagents	behavior can become hard to trace after several layers of hooks and agents
Codex CLI	strong terminal benchmark evidence, sandboxing, approvals, scripting, and implement–test loops	local, scripted, and cloud modes may not share the same permission behavior
Gemini CLI	open-source CLI, large-context models, MCP, extensions, and a natural fit with Google services	preview models, quotas, and very large context can change both cost and consistency
GitHub Copilot CLI	GitHub-native identity, repositories, issues, pull requests, custom agents, skills, and MCP	its strongest value depends on the GitHub and Copilot ecosystem
OpenCode	one terminal interface for several providers and local models	plugins, model aliases, and provider settings make two installations behave differently
Aider	Git-native commits, repository maps, and explicit edit and architect workflows	it is a focused pair-programming tool rather than a general programmable harness
Cursor	low-friction code search, editing, diff review, and background work inside the editor	benchmark scores say little about indexing, privacy settings, or review ergonomics
Pi	small core, multiple providers, branchable sessions, and enough extension points to build a custom harness	the user owns isolation, permissions, extension review, and many team features

The eight tools do not need the same role. Cursor can be the best place to review a diff while Codex or Claude Code does the terminal work. Pi is useful when the experiment itself is the harness. OpenCode emphasizes provider freedom, Copilot CLI minimizes switching for GitHub-centered teams, and Aider keeps the workflow close to Git.

How I would choose today

• For terminal-first repository work, start with Codex CLI or Claude Code.
• For an IDE-centered review loop, start with Cursor.
• For a Google stack or very large source survey, try Gemini CLI.
• For provider switching, try OpenCode.
• For a GitHub-centered workflow, try GitHub Copilot CLI.
• For Git-first terminal pairing, try Aider.
• For harness experiments and controlled model comparisons, use Pi.

Then keep the one that fails least expensively on the actual repository. Familiarity matters: a slightly weaker tool with predictable permissions and recovery can be more useful than a benchmark leader that does not fit the workflow.

A personal replay that would change the answer

Give each serious candidate the same clean repository copy, budget, permissions, and tasks:

Task	What to record
fix an unfamiliar defect	test result, time to locate, unrelated edits
make a cross-file refactor	API preservation, diff size, rollback effort
recover after a failed attempt	repeated mistakes and second-attempt result
handle a permission trap	out-of-scope reads or writes and whether approval worked
resume an interrupted session	lost state and recovery time
run non-interactively	exit code, logs, timeout, and cost cap

Keep task completion, wall time, model/token cost, human interventions, dangerous actions, and cleanup effort. Pass rate alone hides the part that usually determines whether a tool is pleasant to use.

My current shortlist can change quickly because all of these products move quickly. The replay protocol should change much less. That is the durable part of this note.

My reading · Opened notes are remembered in this browser.