Aider Alternatives for AI Pair Programming in the Terminal

The main Aider alternatives for AI pair programming in the terminal are OpenAI's Codex CLI, Anthropic's Claude Code, Google's Gemini CLI, and the open-source agents OpenCode, Qwen Code, goose, Cline, Kilo Code and Plandex. They differ in licence, model choice, approval prompts and how they treat git. Choose on those facts, then scan what changed before you merge.

Safeguard is also an alternative you can switch to without giving up what you already pay for. Bring the Claude, ChatGPT or Codex plan you already pay for and use it inside the Safeguard CLI. Safeguard adds no extra charge for that usage; your plan's own usage limits and terms still apply. You can also connect Claude or ChatGPT to the Safeguard MCP server and use Safeguard from the chat you already use.

Every fact about a third-party tool below comes from that tool's own repository, docs or package page, linked under Sources. There are no rankings here.

What Aider does, and what to check in an alternative

Aider describes itself as "AI pair programming in your terminal." It is licensed Apache-2.0 and installs with two commands:

python -m pip install aider-install
aider-install

You then change into a project and start it with a model and key, for example aider --model sonnet --api-key anthropic=YOUR_KEY. The README says it can connect to almost any LLM, including local models, and its LLM docs list OpenAI, Anthropic, Gemini, Groq, LM Studio, xAI, Azure, Cohere, DeepSeek, Ollama, OpenRouter, GitHub Copilot, Vertex AI, Amazon Bedrock and any OpenAI-compatible API.

Four things define how Aider feels in daily use. Any alternative should be checked against each one.

Git is the undo button

Aider commits each edit with a descriptive message, by default in Conventional Commits style. Before it edits a file that already has uncommitted changes, it commits your work first so the two never mix. /diff shows what changed since your last message, /undo reverts the last commit Aider made, and /commit commits outstanding changes. Commits Aider authored get "(aider)" added to the author and committer names. You can turn all of this off with --no-auto-commits, --no-dirty-commits or --no-git.

Most agents do not work this way. If you move, you need your own checkpoint habit.

Chat modes

Aider has four chat modes. code is the default and edits files. ask discusses your code but never changes it. architect has one model propose a change and an editor model turn it into file edits. help answers questions about Aider itself. Switch with /chat-mode architect, /ask or /code, or start with --architect.

Nearly every alternative has a read-only planning mode that maps to ask. Few have a separate architect and editor model split.

You choose what the model sees

/add puts files in the chat for editing, /read-only adds reference files, and /drop frees context. A repo map, which /map prints, gives the model a view of the wider codebase, and --map-tokens 0 turns it off. Agents that search the repo themselves need less of this manual curation, and give you less control over it.

Lint, test and shell commands

--auto-lint is on by default and --auto-test is off. /run runs a shell command and can add its output to the chat, and /test adds output only when the command fails. Aider can also suggest shell commands, which is on by default, and --yes-always answers yes to every confirmation prompt.

Aider alternatives, tool by tool

Each entry covers what the tool is, how it runs, its licence, and one thing to know before adopting it. Open-source licences were confirmed on each project's own repository page.

OpenAI Codex CLI

Codex CLI is a coding agent from OpenAI that runs locally in your terminal, under the Apache-2.0 licence. Install with npm install -g @openai/codex, brew install --cask codex or the install scripts on the repository page, then run codex and choose Sign in with ChatGPT; the repository recommends this for Plus, Pro, Business, Edu and Enterprise plans, and an API key also works. OpenAI also offers IDE extensions, a desktop app and Codex Web. Know the sandbox before you start: modes are read-only, workspace-write and danger-full-access, network access is off by default, and /permissions switches mode mid-session.

Anthropic Claude Code

Claude Code is "an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools," available in the terminal, IDE, desktop app and browser. It is a commercial product, not open source. Install on macOS, Linux and WSL with curl -fsSL https://claude.ai/install.sh | bash; Claude's pricing page says "Claude Code is included in all paid plans" and it is not part of Free, and a Console account or a supported cloud provider also works. One thing to know: from v2.1.283, auto mode, where a classifier model reviews actions instead of you, is the starting permission mode for interactive terminal sessions, and Shift+Tab switches to Manual (default) or plan.

Google Gemini CLI

Gemini CLI is "an open-source AI agent that brings the power of Gemini directly into your terminal," licensed Apache-2.0, installed with npm install -g @google/gemini-cli or brew install gemini-cli. You sign in with a Google account, a Gemini API key or Vertex AI. Its configuration reference lists default, auto_edit and plan approval modes and says YOLO can only be enabled from the command line. Read the notice first: the README says Gemini CLI was replaced by Antigravity CLI on 18 June 2026 for unpaid tier and Google One users.

OpenCode

OpenCode calls itself "the open source coding agent." It runs as a terminal UI, with a desktop app in beta, under the MIT licence; install with curl -fsSL https://opencode.ai/v2/install | bash or npm i -g @opencode/cli@latest. Its providers page lists 75 or more LLM providers plus local models through Ollama, llama.cpp and LM Studio, connected with /connect. Tab switches between the build agent, which has full access, and plan, which denies file edits and asks before bash commands. More in our OpenCode alternatives guide.

Qwen Code

Qwen Code is "an open-source AI coding agent that lives in your terminal," licensed Apache-2.0. It started from Gemini CLI v0.8.2 and has developed independently since its v0.1. /auth sets the provider, with OpenAI, Anthropic, Gemini and Qwen APIs plus local models through Ollama or vLLM. Approval modes run from Plan through Ask Permissions, Auto-Edit and Auto to YOLO, and the docs page names two different starting modes, so check the mode bar on first run.

goose

goose is an open-source agent that runs on your machine, for research, writing and data work as well as code. It has a CLI, a desktop app for macOS, Linux and Windows, and an API, under Apache-2.0, and is part of the Agentic AI Foundation at the Linux Foundation. The CLI installs with the download_cli.sh script from its releases page. Know the default: goose's permission docs say autonomous mode, which can modify and delete files without approval, is applied by default, and /mode approve switches to Manual Approval.

Cline

Cline describes itself as an open-source coding agent for the IDE, terminal and desktop, licensed Apache-2.0, though its JetBrains plugin is not open source. The CLI installs with npm i -g cline and offers interactive chat or a headless mode with --json output for scripts. It supports Anthropic, OpenAI, Google, OpenRouter, Bedrock, Vertex, Ollama, LM Studio and any OpenAI-compatible API. By default, "every file edit and terminal command requires your approval."

Kilo Code

Kilo Code is "the open source coding agent for building with AI in VS Code, JetBrains, or the CLI," under the MIT licence. The CLI is a fork of OpenCode and installs with npm install -g @kilocode/cli or curl -fsSL https://kilo.ai/cli/install | bash. The README says you can pick from 500 or more models. For CI, kilo run --auto auto-approves permission prompts unless a rule denies the action, and the README says to use it only in trusted environments.

Plandex

Plandex is an "open source AI coding agent" for the terminal, "designed for large projects and real world tasks," under the MIT licence. It installs with curl -sL https://plandex.ai/install.sh | bash, and Windows is supported only through WSL. It works with Anthropic, OpenAI, Google and open-source providers, or an OpenRouter key. Know this before adopting: the README says Plandex Cloud is "winding down as of 10/3/2025 and no longer accepting new users," so plan on running it yourself.

Comparison table

Facts only, from each project's own pages. "Your machine" means the agent edits files locally; your model provider still receives what the agent sends it.

Moving from Aider: setup, worked examples and mistakes

Step by step setup

This walkthrough works for any of the tools above. It assumes a git repository and a clean working tree.

1. Branch first. git switch -c agent-trial. Aider made a commit per edit for you; most alternatives will not, so the branch is your safety net.
2. Take a vulnerability baseline. Run safeguard scan --dir . and keep a copy of the result with cp output/vuln.json baseline-vuln.json. Anything new later came from the trial.
3. Install the new tool with the vendor's own command from the list above, and sign in with an account your organisation approves.
4. Set the strictest useful mode before the first prompt. Codex read-only, Claude Code Manual, Gemini CLI default, OpenCode plan, Qwen Code Ask Permissions, goose /mode approve, or Safeguard Ask.
5. Move your conventions. If you used aider --read CONVENTIONS.md, find which instructions file the new tool reads. Claude Code reads CLAUDE.md, and its docs say it can read an existing AGENTS.md instead.
6. Run one small task and review it. Read git diff, run your tests, and commit the parts you keep with your own message.
7. Scan again and compare. safeguard scan --dir . then diff baseline-vuln.json output/vuln.json.
8. Loosen the mode only after a few clean sessions, and never on a repository you cannot afford to rebuild.

Worked example: mapping Aider habits to other tools

Here is how common Aider habits translate, using only commands each project documents.

A typical session in the new tool then looks like this:

git switch -c fix-date-parsing
safeguard scan --dir .
cp output/vuln.json baseline-vuln.json
claude --permission-mode plan
# review the plan, approve it, review each edit
npm test
git add -p && git commit -m "fix: handle empty date strings"
safeguard scan --dir .
diff baseline-vuln.json output/vuln.json

Swap claude --permission-mode plan for codex, opencode, qwen or safeguard and the surrounding steps stay the same. The scans before and after tell you whether a dependency the agent added brought known vulnerabilities with it.

Common mistakes when leaving Aider

• Assuming the new tool commits for you. Aider's per-edit commits are unusual. Without them, one long session can leave a large uncommitted diff with no easy undo.
• Starting in the most permissive mode. Codex danger-full-access, Claude Code bypassPermissions, Qwen Code YOLO, Gemini CLI YOLO, Kilo run --auto and Safeguard Yolo all skip the prompts. Claude's docs reserve bypass mode for isolated containers and VMs. Treat the others the same way.
• Missing a default. goose starts autonomous and Cline starts by asking about everything. Check before your first prompt, not after.
• Connecting a subscription against its terms. OpenCode's providers page says plugins existed to use Claude Pro or Max models in OpenCode and that "Anthropic explicitly prohibits this." Read your provider's terms before you plug a consumer plan into any tool.
• Skipping the maintenance check. Look at release dates and notices, such as Aider's release gap, Gemini CLI's Antigravity notice and Plandex Cloud's wind-down.
• Reviewing only the lines that changed. A lockfile diff can hide a new transitive dependency. Scan it.

Do the work properly with the Safeguard CLI

If you are moving off Aider, Safeguard is an alternative that keeps what you already pay for. Bring the Claude, ChatGPT or Codex plan you already pay for and use it inside the Safeguard CLI. Safeguard adds no extra charge for that usage; your plan's own usage limits and terms still apply. /model shows or switches the AI model; see the CLI docs for connecting a model.

The Safeguard CLI scans code, container images and SBOMs on your own machine, sends only the results to the Safeguard platform, and includes an assistant that writes code and fixes what it finds. That is the difference from a general coding agent: the assistant also knows your vulnerabilities.

Install

On macOS and Linux:

curl -fsSL https://cli.safeguard.sh/install | bash

On Windows, in PowerShell:

irm https://cli.safeguard.sh/install.ps1 | iex

It is a single static binary with no Node, Python or JVM needed, for macOS, Linux and Windows on x86_64 or arm64 (Windows is x86_64 only). If you came from Aider, there is no Python environment to manage.

First session

Run safeguard with no arguments to open the interactive terminal. Type /login to sign in through the browser, then /scan . to scan the folder you are in. After that, ask in plain language:

how many critical vulns do we have
fix the lodash CVE in this repo

Type @ to attach a file from the folder you started in, much like /add in Aider. /resume picks up the last session in that folder.

Writing and fixing code on your machine

The assistant has tools for your local files and for your Safeguard data, and picks whichever the question needs. It can write files, edit code and run commands, with your approval, so it also works as a general terminal assistant. The docs give "what's the latest file in ~/Downloads" as an example next to the security questions. Our post on the Safeguard CLI as a security-aware coding assistant goes further.

This demo attaches a computer running the Safeguard CLI and a folder, asks for a fix, reviews the diff and approves the command before it runs. The demo notes that computer work from Search is available and expanding.

Click through it at your own pace, or open it full screen.

Using your Claude or ChatGPT plan

Two routes, and you can use both. Inside the CLI, bring the plan you already pay for and pick it with /model. Outside it, connect Claude or ChatGPT to the Safeguard MCP server, so the chat you already use can ask about your projects and findings. This demo connects ChatGPT to the Safeguard MCP server with OAuth and chooses which tools ChatGPT can use.

Click through it at your own pace, or open it full screen.

Approvals

Tools that change things ask before they act. /permissions sets how often: Ask prompts before every action, Auto approves actions it recognises and asks about anything new (the default), and Yolo approves everything. Aider users who liked confirming each step should start in Ask.

Scan what any agent changed

Keep your other tools if you like them. Scan after them:

safeguard scan --dir .
safeguard scan --repo https://github.com/your-org/your-repo
safeguard scan --image your-image:latest
safeguard scan --sbom ./sbom.cdx.json

Each builds an SBOM and matches it against vulnerability databases, writing JSON files such as vuln.json and license.json under ./output. safeguard validate checks CycloneDX, SPDX and VEX documents. When the local runner runs platform scan jobs, your source code never leaves the machine; only SBOMs and findings are uploaded. Policies and gates are evaluated in the platform, which also offers reachability analysis and autonomous remediation with fix PRs (off by default). Our guide to checking what a coding agent changed covers this workflow step by step.

CI with safeguard exec

Aider's -m flag ran one message and exited. The Safeguard equivalent for CI is one non-interactive assistant turn:

safeguard exec "summarise the critical findings in this repository"

One platform on every surface

Start in the terminal, continue in the desktop app for Windows, macOS and Linux, where you ask your supply chain anything in plain English in a native window, and check findings on your phone with the iOS and Android apps. There is an IDE extension for VS Code, JetBrains and Open VSX editors, and a browser extension for Chrome. With nothing to install, there is a Cloud CLI on Safeguard's infrastructure, a cloud browser, and a cloud sandbox tenant against a seeded demo repo, free for seven days with no card.

Download Safeguard, then run /scan . in the repository you work in today.

FAQ

Is Aider still maintained?

Its PyPI page lists version 0.86.2, released on 12 February 2026, with the previous release on 13 August 2025. Check the PyPI page and the repository for the latest state before a team rollout.

Which Aider alternatives are open source?

Codex CLI (Apache-2.0), Gemini CLI (Apache-2.0), OpenCode (MIT), Qwen Code (Apache-2.0), goose (Apache-2.0), Cline (Apache-2.0, except its JetBrains plugin), Kilo Code (MIT) and Plandex (MIT). Claude Code is a commercial product.

Which alternative commits every change to git like Aider does?

None of the project pages we checked describe per-edit automatic commits the way Aider's git docs do. Claude Code's docs say it stages changes, writes commit messages and opens pull requests when asked. Plan on your own branch and commit routine.

Can I switch to Safeguard and keep my Claude, ChatGPT or Codex plan?

Yes. Bring the Claude, ChatGPT or Codex plan you already pay for and use it inside the Safeguard CLI. Safeguard adds no extra charge for that usage; your plan's own usage limits and terms still apply. You can also connect Claude or ChatGPT to the Safeguard MCP server and use Safeguard from that chat.

Can I use my ChatGPT plan in a terminal coding agent?

Codex CLI signs in with ChatGPT and recommends that for Plus, Pro, Business, Edu and Enterprise plans. OpenCode's providers page also offers a ChatGPT Plus/Pro sign-in through /connect. And you can bring your ChatGPT plan to the Safeguard CLI.

Which alternatives work with local models?

Aider, OpenCode, Qwen Code, goose and Cline all document local options such as Ollama, LM Studio or vLLM on their own pages.

Do these tools ask before running commands?

It depends on the tool and mode. Cline asks about every edit and command by default. goose starts autonomous. Codex combines a sandbox with an approval policy. The Safeguard CLI asks according to Ask, Auto or Yolo, with Auto as the default.

Does my code leave my machine?

Terminal agents edit files locally, but your model provider receives whatever the agent sends it. Codex Web, Claude Code on the web and Kilo's Cloud Agent run in the vendor's cloud. With the Safeguard local runner, source code stays on your machine and only SBOMs and findings are uploaded.

Try it yourself

This walkthrough sets up the runner on your own machine: pick the platform, start it, approve it with its device code, then use the same CLI to scan from the terminal.

Click through it at your own pace, or open it full screen.

Quick check

Sources

• Aider repository
• aider-chat on PyPI
• Aider chat modes
• Aider git integration
• Aider in-chat commands
• Aider options reference
• Aider LLM connections
• OpenAI Codex repository
• Codex approvals and security
• Claude Code overview
• Claude Code permission modes
• Claude pricing
• Gemini CLI repository
• Gemini CLI configuration reference
• OpenCode repository
• OpenCode providers docs
• Qwen Code repository
• Qwen Code approval modes
• goose repository
• goose permission modes
• Cline repository
• Kilo Code repository
• Plandex repository
• Safeguard CLI docs
• Safeguard CLI interactive terminal docs
• Safeguard download page

Last reviewed: 21 October 2026.