Aider Integration

Configure Aider with QCode.cc: Claude over the Anthropic endpoint (anthropic/ prefix), GPT and Chinese models over the OpenAI-compatible endpoint

Updated 2026-09-18
On This Page

Last verified: 2026-09-18 · 📄 Per official docs (upstream repository; see the status note below)

Upstream activity: the last commit in Aider-AI/aider is dated 2026-05-22 with no releases since (checked via the GitHub API on 2026-09-18). Everything on this page still works; for an actively maintained alternative look at Cline or Kilo Code.

At a glance

Item Details
Models you can use Claude ✅ (anthropic/ prefix on the Anthropic endpoint) · GPT ✅ · Chinese models ✅ (OpenAI-compatible endpoint) · Gemini ❌ (no Gemini route documented here)
Protocol & Base URL Anthropic: https://api.qcode.cc/api · OpenAI: https://api.qcode.cc/openai/v1
Relevant env vars ANTHROPIC_API_BASE and OPENAI_API_BASE point to the two legs respectively
Where to configure CLI flags or ~/.aider.conf.yml
Official docs aider.chat

Aider is a popular open-source AI pair programmer (39K+ GitHub stars) that runs in your terminal and supports 100+ languages. It routes models through LiteLLM, so it speaks both the Anthropic and the OpenAI protocol.

🔴 Read this first: Claude requires the Anthropic endpoint

QCode's OpenAI-compatible endpoint does not accept Claude models. Pointing OPENAI_API_BASE at QCode and then using an openai/claude-… model name returns model_not_available_on_endpoint.

Model you want Aider model prefix Env var Value
Claude anthropic/ ANTHROPIC_API_BASE https://api.qcode.cc/api
GPT / the four Chinese families openai/ OPENAI_API_BASE https://api.qcode.cc/openai/v1

See Endpoints & API Paths.

Why Aider

  • Fully open source: you only pay for API usage
  • Architect mode: one model plans, another edits — better output quality
  • Deep Git integration: every AI edit becomes a git commit
  • Repository Map: tree-sitter based indexing of the whole codebase
  • Both protocols: Claude over Anthropic, GPT / Chinese models over OpenAI

Install

# pipx recommended (isolated install)
pipx install aider-chat

# or pip
pip install aider-chat

Configure Claude (Anthropic endpoint)

export ANTHROPIC_API_BASE="https://api.qcode.cc/api"
export ANTHROPIC_API_KEY="cr_your_qcode_key"

aider --model anthropic/claude-sonnet-5

LiteLLM appends /v1/messages to this base automatically, so stop at /api with no trailing slash.

The variable name varies by version: LiteLLM has historically used both ANTHROPIC_API_BASE and ANTHROPIC_BASE_URL. If one has no effect, try the other, or pass --anthropic-api-key on the command line. Defer to the Aider docs.

Persist it (append to ~/.zshrc or ~/.bashrc):

echo 'export ANTHROPIC_API_BASE="https://api.qcode.cc/api"' >> ~/.zshrc
echo 'export ANTHROPIC_API_KEY="cr_your_qcode_key"' >> ~/.zshrc
source ~/.zshrc

Configure GPT and Chinese models (OpenAI-compatible endpoint)

export OPENAI_API_BASE="https://api.qcode.cc/openai/v1"
export OPENAI_API_KEY="cr_your_qcode_key"

aider --model openai/gpt-5.5

The four Chinese families (glm-5.2 / kimi-k3 / deepseek-v4-pro / qwen3.7-max …) work on both legs; ids are in Chinese Models.

Usage

cd /path/to/your/project

# Everyday driver
aider --model anthropic/claude-sonnet-5

# Flagship
aider --model anthropic/claude-opus-5

One model plans, another applies the edits:

# Opus plans + Sonnet edits (recommended)
aider --architect --model anthropic/claude-opus-5 --editor-model anthropic/claude-sonnet-5

# Sonnet plans + Haiku edits (cheaper)
aider --architect --model anthropic/claude-sonnet-5 --editor-model anthropic/claude-haiku-4-5

Both models must sit on the same protocol leg. Mixing them (e.g. anthropic/ planner with an openai/claude-… editor) fails on the editor side.

Common commands

Inside an Aider session:

Command What it does
/add file.py Add a file to the chat context
/drop file.py Remove a file
/run pytest Run a command and send the output to the AI
/diff Show all changes
/undo Undo the last AI edit
/commit Commit current changes
/help Show help

Alternative endpoints

The four access domains are functionally identical and differ only in network routing. One key works on all of them:

Node Anthropic (Claude) OpenAI (GPT / Chinese)
Global https://api.qcode.cc/api https://api.qcode.cc/openai/v1
Asia (best from mainland China) https://asia.qcode.cc/api https://asia.qcode.cc/openai/v1
US https://us.qcode.cc/api https://us.qcode.cc/openai/v1
Europe https://eu.qcode.cc/api https://eu.qcode.cc/openai/v1

Available models

Model Name in Aider Notes
Claude Sonnet 5 anthropic/claude-sonnet-5 Recommended, best value
Claude Opus 5 anthropic/claude-opus-5 Most capable
Claude Haiku 4.5 anthropic/claude-haiku-4-5 Cheap and fast
GPT 5.5 openai/gpt-5.5 OpenAI flagship
GLM 5.2 openai/glm-5.2 Chinese, low unit price

4.x models such as claude-sonnet-4-6 and claude-opus-4-8 are still sold (same anthropic/ prefix). The live list is on qcode.cc/models.

Aider vs Claude Code CLI

Dimension Aider Claude Code CLI
Open source Fully open Closed
Git integration Auto-commit per edit Manual /commit
Architect mode Two-model plan + edit Single model
Tooling File edit + shell Richer (LSP, search, browser)
Context handling Repository Map indexing 200K–1M token window
Quota Shares your QCode.cc plan Shares your QCode.cc plan

A good pairing: Aider for quick edits and Architect-mode planning, Claude Code CLI for deep project analysis and automation.

Troubleshooting

model_not_available_on_endpoint

You sent a Claude model to the OpenAI leg. Check two things: the model prefix should be anthropic/, not openai/; and the base should be ANTHROPIC_API_BASE=https://api.qcode.cc/api.

"Model not found"

Aider needs the prefix to know which provider to use:

# correct
aider --model anthropic/claude-sonnet-5

# wrong (missing prefix)
aider --model claude-sonnet-5

Requests time out

Try another node, or raise the timeout:

aider --model anthropic/claude-sonnet-5 --timeout 120

Next

Related Documents

Roo Code Setup
Use QCode.cc from the Roo Code VS Code extension: pick the Anthropic provider, tick the custom base URL, and Claude works
Connect SillyTavern to QCode
Chat with QCode.cc's Claude / GPT models in SillyTavern; an honest note on whether gpt-image-2 image generation can be connected, plus alternatives
Droid (Factory) Setup
Add QCode.cc as a BYOK custom model in Factory's Droid CLI: provider=anthropic plus baseUrl in ~/.factory/settings.json
🚀
Get Started with QCode — Claude Code & Codex
One plan for both Claude Code and Codex, Asia-Pacific low latency
View Pricing Plans → Create Account
Team of 3+?
Enterprise: dedicated domain + sub-key management + ban protection, from ¥250/person/mo
Learn Enterprise →