Aider · Pair-programming AI in your terminal

Image and voice generation, called from inside Aider sessions.

Aider is a terminal-first pair programmer that has stayed lean and fast. It does not ship with image generation or TTS, and the popular pattern is to bolt those on as external commands. Gathos slots in as a Bash-callable tool so an Aider session can generate assets without leaving the terminal.

Install
curl -sL https://gathos.com/install.sh | bash

A sample session

You
aider> i need an OG image for the new /pricing route. 1200×630, our brand colors, "Flat $18/month" headline.
Aider
Got it. Running `gathos image-gen --width 1200 --height 630 --prompt "Cream background with matcha-green accents, bold serif text Flat dollar 18 slash month, minimalist Clay aesthetic"`...
Aider
✓ saved public/og-pricing.png. Adding it to your <Helmet /> in src/pages/Pricing.jsx.
You
aider> generate a 30-second voiceover of the page summary in my cloned voice for a launch tweet.
Aider
Running `gathos tts --voice my-clone --text "Gathos is..."`. Output: launch-vo.mp3 (28s).

Why pair Gathos with Aider

01
Aider stays lean · Gathos plugs in via Bash
Aider does not have a plugin system in the modern agent-skill sense. Gathos integrates as a CLI binary that Aider can shell out to. Same flexibility, no new abstraction.
02
Bundled image + voice means one auth, one bill
Aider already orchestrates several tools per session. Adding Gathos means one more API key, not two · image generation and TTS share a single key.
03
Predictable cost is critical for terminal pair-programming
Aider is good at iterating in tight loops. Per-image pricing turns a 30-minute session into an unintended $20 bill. Gathos is flat · generate as many drafts as the work needs.
04
Skills are markdown, not Python plugins
The Gathos skills are single-file recipes that document what the agent should call when. Aider can read them as context (e.g., `aider --read .gathos/skills/youtube-thumbnails.md`) and follow the recipe.

Frequently asked questions

Does Aider have a "skills" system like Claude Code or Cursor?+
Not in the same form factor. Aider stays close to the terminal-pair-programming metaphor. Gathos works by adding a CLI tool that Aider can call directly via shell-out, plus markdown recipes Aider can read with `--read`.
Will Aider always know to call Gathos for an image task?+
Yes if you `--read` the relevant skill markdown, or if you tell it once at the start of the session. Aider is good at remembering tool conventions within a session.
Is the CLI binary required?+
No. You can also call the Gathos REST API directly from a one-line `curl` and Aider handles that fine. The CLI is a convenience.
Does this work with Aider in a Docker container?+
Yes. Mount your Gathos config volume into the container and the CLI works the same. Many teams keep Aider in a dev container.
What about cost? Aider is loop-friendly.+
Flat $18/month is the only safe answer to a loop-friendly agent. Per-image APIs and Aider together produce surprise bills more often than not.

Using a different agent?

GitHub Copilot

Image and voice generation, plugged into GitHub Copilot.

GitHub Copilot supports custom agents and MCP servers in 2026, which means Gathos can be wired in as a callable tool inside any IDE that ships Copilot. The setup is one curl line and one config entry.

ChatGPT Custom GPTs

Image and voice generation, callable from a custom ChatGPT GPT.

Custom ChatGPT GPTs support Actions, which are HTTPS-callable tools the GPT can invoke during a conversation. Gathos exposes a clean OpenAPI schema so wiring it as an Action takes about ten minutes.

Continue

Image and voice generation in Continue.dev sessions.

Continue is the open-source alternative to Cursor, focused on developer control and self-hosting. It supports custom slash commands and tools through a clean config. Gathos plugs in as a callable tool so the agent can generate assets from the same editor session.

Claude Code

Image generation and voice cloning for Claude Code.

Claude Code is the fastest-growing agent for power users. Gathos is the unified image + TTS API that plugs into Claude Code with a single install command. This page shows the install, a sample session, and the pre-built skills most teams install first.

Gemini CLI

Image and voice generation, native to Gemini CLI.

Gemini CLI's Skills spec went GA in early 2026 · same single-file markdown convention Anthropic and Codeium adopted. Gathos installs as a Gemini CLI skill so the agent can generate images and voiceover from any terminal session.

Cursor

Image generation and voice cloning in Cursor.

Cursor is where millions of developers write code every day. Gathos plugs into Cursor as a set of agent-callable skills · you describe what you want in a chat message, Cursor runs the API, and the output lands in your project folder. Flat $18/month.

Windsurf

Image generation and voice cloning, native to Windsurf.

Windsurf shipped Cascade Skills in Wave 13 (January 2026), and the spec is the same single-file markdown convention Anthropic and Cursor adopted. Gathos plugs in as a Cascade Skill so the agent can generate thumbnails, product shots, voiceover, and dubbed videos without you leaving the editor.

Plug Gathos into Aider in one command.