An Anthropic-compatible proxy that lets Claude Code, the VS Code extension, JetBrains ACP, and chat bots route traffic through alternative model providers.
Free Claude Code forwards Anthropic Messages API traffic from Claude Code to a configurable set of upstream providers, including NVIDIA NIM, Kimi, Wafer, OpenRouter, DeepSeek, LM Studio, llama.cpp, Ollama, OpenCode Zen, and Z.ai. The client-side protocol that Claude Code expects remains stable, while the operator retains full control over which model — free, paid, or local — handles each request.
Quick Start · Providers · Clients · Integrations · Development
Free Claude Code provides:
- A drop-in proxy for the Anthropic API calls issued by Claude Code.
- Ten provider backends: NVIDIA NIM, Kimi, Wafer, OpenRouter, DeepSeek, LM Studio, llama.cpp, Ollama, OpenCode Zen, and Z.ai.
- Per-tier model routing, enabling Opus, Sonnet, Haiku, and fallback traffic to be directed to different providers.
- Native support for the Claude Code
/modelpicker via the proxy's/v1/modelsendpoint. Claude Code must opt in to Gateway model discovery; see Model Picker. - Streaming, tool use, reasoning and thinking block handling, and local request optimizations.
- An optional Discord or Telegram bot wrapper for remote coding sessions.
- Optional usage through the VS Code extension.
- Optional voice-note transcription via local Whisper or NVIDIA NIM.
- A local administrative interface at
/adminfor editing supported proxy settings, validating changes, and verifying providers. Access is restricted to loopback.
Click to expand or collapse
The repository ships one-shot installers that perform steps 1, 2, and 4 below (Node.js + Claude Code CLI, uv + Python 3.14, and the proxy itself). Step 3 — obtaining the NVIDIA NIM API key — remains manual.
Clone the repository first, then run the script for your platform. Both scripts are idempotent and re-run safe (they upgrade in place).
macOS and Linux:
git clone https://github.com/Pushpenderrathore/claude-code.git
cd claude-code
chmod +x install.sh
./install.shWindows PowerShell:
git clone https://github.com/Pushpenderrathore/claude-code.git
cd claude-code
powershell -ExecutionPolicy Bypass -File .\install.ps1The script will:
- Install Node.js + npm (via
brew/apt/dnf/pacmanon Unix, orwinget/chocoon Windows) if missing. - Install the Claude Code CLI (
npm i -g @anthropic-ai/claude-code). - Install or update uv, then install Python 3.14.
- Install the proxy (
fcc-server,fcc-claude,fcc-init) as a uv tool.
Skip ahead to 3. Obtain an NVIDIA NIM API Key once the script finishes.
If you prefer to do it step by step, install the latest release of Claude Code:
npm install -g @anthropic-ai/claude-codeInstall the latest version of uv along with Python 3.14.
macOS and Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh
uv self update
uv python install 3.14Windows PowerShell:
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
uv self update
uv python install 3.14Create a free NVIDIA NIM API key and keep it available for the Admin UI configuration step. See NVIDIA NIM provider setup for details.
uv tool install --force git+https://github.com/Pushpenderrathore/claude-code.gitThe same command is used to update an existing installation.
fcc-serverAfter startup, Uvicorn prints the proxy bind address and the application logs the administrative URL:
INFO: Admin UI: http://127.0.0.1:8082/admin (local-only)
Most terminals render this URL as a clickable link. Substitute the value of PORT if it differs from 8082.
Open the Admin UI URL printed in the terminal.
Paste the NVIDIA NIM API key into NVIDIA_NIM_API_KEY, click Validate, and then click Apply.
The default model is preset to nvidia_nim/z-ai/glm4.7. This may be changed at any time through the same Admin UI.
fcc-claudefcc-claude reads the currently configured port and authentication token on each invocation, sets the required Claude Code environment variables (including a 190,000-token CLAUDE_CODE_AUTO_COMPACT_WINDOW value for auto-compaction), and then launches the underlying claude command.
The following issues are commonly encountered while completing the Quick Start. Each entry pairs a symptom with the resolution.
The uv installer places the binary in $HOME/.local/bin, but the current shell session does not yet include that directory in PATH. The installer prints this on completion; if it has scrolled out of view, apply the fix below.
macOS and Linux (bash or zsh):
source $HOME/.local/bin/env
uv --versionAfter verifying that uv --version succeeds, continue with uv python install 3.14.
To make the change permanent for new shell sessions, append the same source line to the appropriate shell rc file:
# zsh (default on macOS)
echo 'source $HOME/.local/bin/env' >> ~/.zshrc
source ~/.zshrc
# bash
echo 'source $HOME/.local/bin/env' >> ~/.bashrc
source ~/.bashrc
# fish
echo 'source $HOME/.local/bin/env.fish' >> ~/.config/fish/config.fishOn Windows PowerShell, close and reopen the terminal after the installer completes, or run refreshenv if the Chocolatey helpers are available.
uv uses the GNU-style long flag. Use uv --version (two dashes) or uv self version.
The Uvicorn entry point is server:app, not service:app or app:app. From a source checkout:
uv run uvicorn server:app --host 0.0.0.0 --port 8082Most users do not need this command — install the packaged proxy and run fcc-server as described in Quick Start instead. Use the uvicorn form only when working from a local clone of the repository.
The directory is only named nvidia-nim when the repository is cloned with that name as the target. Either clone with an explicit target directory, or cd into the default claude-code directory produced by git clone:
# Option A — clone into an explicitly named directory
git clone https://github.com/Pushpenderrathore/claude-code.git nvidia-nim
cd nvidia-nim
# Option B — use the default directory name
git clone https://github.com/Pushpenderrathore/claude-code.git
cd claude-codeThe file is created successfully; macOS Finder and ls hide dotfiles by default. Confirm with ls -la or open the file directly:
ls -la .env
open -e .env # macOS, opens in TextEditgit clone refuses to overwrite a non-empty target. Remove or rename the existing directory if its contents are not needed, or clone into a different target name:
rm -rf nvidia-nim # destructive — verify contents first
git clone https://github.com/Pushpenderrathore/claude-code.git nvidia-nimStop any previously started instance of fcc-server (or any other process bound to the port), or choose a different port:
lsof -i :8082 # identify the process holding the port (macOS / Linux)
fcc-server --port 8090When changing the port, update ANTHROPIC_BASE_URL for any connected client (Claude Code CLI, VS Code, JetBrains ACP) to match.
The repository ships matching uninstall scripts that mirror the installers. By default they remove only the proxy (the thing this repository installs). Shared toolchain (Node.js, npm, uv, uv-managed Pythons) is intentionally left in place — uninstall those manually if you also want them gone.
macOS and Linux:
chmod +x Uninstall.sh
./Uninstall.sh # remove the proxy only
./Uninstall.sh --with-claude-cli # also npm-uninstall @anthropic-ai/claude-code
./Uninstall.sh --purge # also delete ~/.fcc/ (API keys, managed .env, logs)
./Uninstall.sh --all # shorthand for --with-claude-cli --purgeWindows PowerShell:
powershell -ExecutionPolicy Bypass -File .\Uninstall.ps1 # proxy only
powershell -ExecutionPolicy Bypass -File .\Uninstall.ps1 -WithClaudeCli # also npm-uninstall the CLI
powershell -ExecutionPolicy Bypass -File .\Uninstall.ps1 -Purge # also delete %USERPROFILE%\.fcc
powershell -ExecutionPolicy Bypass -File .\Uninstall.ps1 -All # shorthand for bothEach step is idempotent: if a component is already gone, the script reports it and continues.
Note:
--purge/-Purgedeletes~/.fcc/, which contains yourNVIDIA_NIM_API_KEY(and any other provider keys), the managed.env, and proxy logs. Back the directory up first if you intend to reinstall later.
The automated scripts run these commands under the hood. Run them yourself if you want fine-grained control.
# 1. Remove the proxy (fcc-server, fcc-claude, fcc-init, claude-code shim)
uv tool uninstall claude-code
# 2. (Optional) Remove the Claude Code CLI
npm uninstall -g @anthropic-ai/claude-code
# 3. (Optional, destructive) Remove user config — contains API keys
rm -rf ~/.fcc # macOS / Linux
# Remove-Item -LiteralPath "$env:USERPROFILE\.fcc" -Recurse -Force # Windows PowerShell
# 4. (Optional) Remove shared toolchain — only do this if no other project needs it
uv self uninstall # remove uv itself
uv python uninstall 3.14 # remove the uv-managed Python (run before uv self uninstall)Select a provider, supply its API key or local URL through the Admin UI, and set MODEL to a provider-prefixed model slug. MODEL acts as the fallback. The values MODEL_OPUS, MODEL_SONNET, and MODEL_HAIKU may be used to override routing for individual Claude Code model tiers.
1. NVIDIA NIM
Obtain an API key at build.nvidia.com/settings/api-keys.
In the Admin UI, paste the key into NVIDIA_NIM_API_KEY. The default value of MODEL is nvidia_nim/z-ai/glm4.7.
Common examples:
nvidia_nim/z-ai/glm4.7nvidia_nim/z-ai/glm5nvidia_nim/moonshotai/kimi-k2.5nvidia_nim/minimaxai/minimax-m2.5
Available models can be browsed at build.nvidia.com.
2. Kimi
Obtain an API key at platform.moonshot.ai/console/api-keys.
In the Admin UI, paste the key into KIMI_API_KEY and set MODEL to a Kimi slug, such as kimi/kimi-k2.5.
Available models can be browsed at platform.moonshot.ai.
3. Wafer
Obtain an API key from wafer.ai. In the Admin UI, paste the key into WAFER_API_KEY and set MODEL to a Wafer Pass model, such as wafer/DeepSeek-V4-Pro.
Common examples:
wafer/DeepSeek-V4-Prowafer/MiniMax-M2.7wafer/Qwen3.5-397B-A17Bwafer/GLM-5.1
This provider connects to Wafer's Anthropic-compatible endpoint at https://pass.wafer.ai/v1/messages.
4. OpenRouter
Obtain an API key at openrouter.ai/keys.
In the Admin UI, paste the key into OPENROUTER_API_KEY and set MODEL to an OpenRouter slug, such as open_router/stepfun/step-3.5-flash:free.
Browse all models or the free models collection.
5. DeepSeek
Obtain an API key at platform.deepseek.com/api_keys.
In the Admin UI, paste the key into DEEPSEEK_API_KEY and set MODEL to a DeepSeek slug, such as deepseek/deepseek-chat.
This provider connects to DeepSeek's Anthropic-compatible endpoint rather than the OpenAI chat-completions endpoint.
6. LM Studio
Start the LM Studio local server and load a model. In the Admin UI, retain or update LM_STUDIO_BASE_URL, then set MODEL to the model identifier shown by LM Studio, prefixed with lmstudio/.
Models with tool-use support are preferred for Claude Code workflows.
7. llama.cpp
Start llama-server with an Anthropic-compatible /v1/messages endpoint and sufficient context to handle Claude Code requests.
In the Admin UI, retain or update LLAMACPP_BASE_URL, then set MODEL to the local model slug, prefixed with llamacpp/.
Context size is significant for local coding models. If llama.cpp returns HTTP 400 for ordinary Claude Code requests, increase --ctx-size and confirm that the selected model and server build support the required features.
8. Ollama
Run Ollama and pull a model:
ollama pull llama3.1
ollama serveIn the Admin UI, retain or update OLLAMA_BASE_URL, then set MODEL to the tag reported by ollama list, prefixed with ollama/.
OLLAMA_BASE_URL is the Ollama server root and should not include a /v1 suffix. Examples of model slugs include ollama/llama3.1 and ollama/llama3.1:8b.
9. OpenCode Zen
Obtain an API key at opencode.ai/auth.
In the Admin UI, paste the key into OPENCODE_API_KEY and set MODEL to an OpenCode Zen model slug, such as opencode/gpt-5.3-codex.
OpenCode Zen is a curated model gateway that provides access to models from Anthropic, OpenAI, Google, DeepSeek, and others through a single API key and an OpenAI-compatible endpoint at https://opencode.ai/zen/v1.
Common examples:
opencode/gpt-5.3-codexopencode/claude-sonnet-4opencode/deepseek-v4-flash-free(free)opencode/gemini-3-flashopencode/big-pickle(free)opencode/glm-5.1
Available models can be browsed at opencode.ai.
10. Z.ai
Obtain an API key at Z.ai/manage-apikey/apikey-list.
In the Admin UI, paste the key into ZAI_API_KEY and set MODEL to a Z.ai model slug, such as zai/glm-5.1.
Z.ai exposes GLM models through the OpenAI-compatible Coding Plan endpoint at https://api.z.ai/api/coding/paas/v4.
Common examples:
zai/glm-5.1zai/glm-5-turbo
Available models can be browsed at Z.ai.
Each model tier may use a distinct provider by configuring MODEL_OPUS, MODEL_SONNET, and MODEL_HAIKU in the Admin UI. Any tier left blank inherits the value of MODEL.
For example, Opus may be routed to nvidia_nim/moonshotai/kimi-k2.5, Sonnet to open_router/deepseek/deepseek-r1-0528:free, Haiku to lmstudio/unsloth/GLM-4.7-Flash-GGUF, while the fallback MODEL remains zai/glm-5.1.
For terminal usage, the installed launcher is recommended:
fcc-claudefcc-server must remain running during use. The Admin UI manages proxy configuration and restarts the server when runtime settings change. fcc-claude reads the current Admin UI-managed port and authentication token on each invocation and sets CLAUDE_CODE_AUTO_COMPACT_WINDOW to 190000 to enable auto-compaction.
Open Settings, search for claude-code.environmentVariables, choose Edit in settings.json, and add the following:
"claudeCode.environmentVariables": [
{ "name": "ANTHROPIC_BASE_URL", "value": "http://localhost:8082" },
{ "name": "ANTHROPIC_AUTH_TOKEN", "value": "freecc" },
{ "name": "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY", "value": "1" },
{ "name": "CLAUDE_CODE_AUTO_COMPACT_WINDOW", "value": "190000" }
]Reload the extension. If the extension presents a login screen, complete the Anthropic Console path once; the local proxy will continue to handle model traffic after the environment variables take effect.
Edit the installed Claude ACP configuration:
- Windows:
C:\Users\%USERNAME%\AppData\Roaming\JetBrains\acp-agents\installed.json - Linux and macOS:
~/.jetbrains/acp.json
Set the environment block for acp.registry.claude-acp:
"env": {
"ANTHROPIC_BASE_URL": "http://localhost:8082",
"ANTHROPIC_AUTH_TOKEN": "freecc",
"CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY": "1",
"CLAUDE_CODE_AUTO_COMPACT_WINDOW": "190000"
}Restart the IDE after modifying the file.
For all integrations described below, managed proxy settings should be modified exclusively through the Admin UI at /admin: edit the relevant fields, click Validate, and then click Apply. The footer indicates where the managed configuration is stored. Manual editing of that file is intentionally not documented here.
The bot wrapper executes Claude Code sessions remotely, streams progress, supports reply-based conversation branches, and provides controls for stopping or clearing tasks.
Discord
- Create a bot in the Discord Developer Portal.
- Enable the Message Content Intent.
- Invite the bot with read, send, and message-history permissions.
- Record the bot token and the numeric identifier of each channel in which the bot should respond.
Telegram
- Create a bot through @BotFather and record the bot token.
- Retrieve the numeric user ID from @userinfobot so that access can be restricted to a single user.
Admin UI Configuration
- With
fcc-serverrunning, open the Admin UI URL from the terminal output. - In the sidebar, select Messaging.
- Set Messaging Platform to either discord or telegram.
- For Discord, paste the Discord Bot Token and Allowed Discord Channels. For Telegram, paste the Telegram Bot Token and Allowed Telegram User ID.
- Set Allowed Directory to an absolute path on the host running the proxy — the workspace root the bot is permitted to use.
- Click Validate, then Apply. Restart the server if prompted.
Admin UI — Messaging view (platform, bots, and voice)
Commands
/stopcancels a task. Replying to a task message limits the cancellation to that branch./clearresets sessions. Replying to a message limits the reset to that branch./statsdisplays session state.
Voice notes are supported on Discord and Telegram once the proxy installation has been extended with the appropriate optional extras. Re-run uv tool install --force with the required extras (the Git URL matches the Quick Start):
# NVIDIA NIM transcription via Riva gRPC
uv tool install --force "claude-code[voice] @ git+https://github.com/Pushpenderrathore/claude-code.git"
# Local Whisper (CPU or CUDA)
uv tool install --force "claude-code[voice_local] @ git+https://github.com/Pushpenderrathore/claude-code.git"
# Both backends
uv tool install --force "claude-code[voice,voice_local] @ git+https://github.com/Pushpenderrathore/claude-code.git"For CUDA-accelerated local Whisper, append --torch-backend cu130 to the voice_local installation command. Restart fcc-server after each reinstallation.
In the Admin UI, open Messaging and scroll to Voice. Enable Voice Notes, choose a Whisper Device (cpu, cuda, or nvidia_nim), set a Whisper Model, and provide a Hugging Face Token if the configuration requires one. For nvidia_nim transcription, install the voice extra and set the NVIDIA NIM API Key on the Providers view. The screenshot above illustrates the Voice block within the same view.
Diagram source: assets/how-it-works.mmd.
Key components:
- FastAPI exposes Anthropic-compatible routes including
/v1/messages,/v1/messages/count_tokens, and/v1/models. - Model routing resolves the requested Claude model name to
MODEL_OPUS,MODEL_SONNET,MODEL_HAIKU, orMODEL. - NVIDIA NIM, OpenCode Zen, and Z.ai use OpenAI chat streaming, which is translated into Anthropic Server-Sent Events.
- Wafer, OpenRouter, DeepSeek, LM Studio, llama.cpp, and Ollama use Anthropic Messages-style transports.
- The proxy normalizes thinking blocks, tool calls, token usage metadata, and provider error responses into the shape expected by Claude Code.
- Request optimizations resolve trivial Claude Code probes locally to reduce latency and conserve quota.
claude-code/
├── server.py # ASGI entry point
├── api/ # FastAPI routes, service layer, routing, optimizations
├── core/ # Shared Anthropic protocol helpers and SSE utilities
├── providers/ # Provider transports, registry, rate limiting
├── messaging/ # Discord and Telegram adapters, sessions, voice
├── cli/ # Package entry points and Claude process management
├── config/ # Settings, provider catalog, logging
└── tests/ # Unit and contract tests
Use the following workflow when developing or running directly from a checkout:
git clone https://github.com/Pushpenderrathore/claude-code.git
cd claude-code
uv run uvicorn server:app --host 0.0.0.0 --port 8082uv run ruff format
uv run ruff check
uv run ty check
uv run pytestThese should be executed in the order shown prior to pushing. Continuous integration enforces the same sequence.
pyproject.toml installs the following entry points:
fcc-server: starts the proxy with the configured host and port.fcc-init: optional advanced scaffold for~/.fcc/.env. The Admin UI is preferred for routine configuration.fcc-claude: launches Claude Code with the configured local proxy URL, authentication token, model-discovery flag, and a 190,000-tokenCLAUDE_CODE_AUTO_COMPACT_WINDOWvalue for auto-compaction.claude-code: a compatibility as forfcc-server.
- Add OpenAI-compatible providers by extending
OpenAIChatTransport. - Add Anthropic Messages providers by extending
AnthropicMessagesTransport. - Register provider metadata in
config.provider_catalogand factory wiring inproviders.registry. - Add messaging platforms by implementing the
MessagingPlatforminterface withinmessaging/.
This section consolidates the end-to-end setup links and commands into a single narrative. It is provided as a beginner-friendly companion to the Quick Start above and references the same scripts, ports, and tokens.
- Install uv (Python toolchain): docs.astral.sh/uv/getting-started/installation
- Free Claude Code repository: github.com/Pushpenderrathore/claude-code
- NVIDIA NIM API keys: build.nvidia.com/settings/api-keys
- OpenRouter API keys: openrouter.ai/keys
- DeepSeek API keys: platform.deepseek.com/api_keys
Install Python and clone the repository:
uv python install 3.14
git clone https://github.com/Pushpenderrathore/claude-code.git nvidia-nim
cd nvidia-nimCopy the environment template.
macOS and Linux:
cp .env.example .envWindows PowerShell:
Copy-Item .env.example .envStart the proxy:
uv run uvicorn server:app --host 0.0.0.0 --port 8082Note: The proxy must remain running in a background terminal for the duration of the Claude Code session.
Install Claude Code:
curl -fsSL https://claude.ai/install.sh | bashLaunch Claude Code against the local proxy.
Windows PowerShell:
$env:ANTHROPIC_AUTH_TOKEN="freecc"; $env:ANTHROPIC_BASE_URL="http://localhost:8082"; claudeBash:
ANTHROPIC_AUTH_TOKEN="freecc" ANTHROPIC_BASE_URL="http://localhost:8082" claudeNote: Both environment variables (
ANTHROPIC_AUTH_TOKENandANTHROPIC_BASE_URL) must be exported on every invocation, or the launcher scriptfcc-claudemay be used instead — see Quick Start step 7.
Most tutorials recommend running local models through Ollama or LM Studio. In practice, the hardware required to run a model that is both capable and responsive is outside the reach of most users — even on a high-end Mac with 64 GB of RAM and an M-series chip, capable local models are noticeably slow, and faster local models trade away the reasoning quality that Claude Code workflows rely on.
Claude Code's official pricing starts at $20/month with usage limits that are restrictive for sustained coding sessions, and even the $100/month tier is constraining for heavy users. The approach documented here routes Claude Code traffic to a free NVIDIA NIM endpoint through this proxy, preserving the Claude Code client experience while removing the cost ceiling.
Install the Claude Code CLI from the official source:
curl -fsSL https://claude.ai/install.sh | bashDo not launch claude yet — proxy configuration must be completed first.
Install uv, update it to the latest release, and provision Python 3.14:
# macOS and Linux
curl -LsSf https://astral.sh/uv/install.sh | sh
uv self update
uv python install 3.14# Windows PowerShell
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
uv self update
uv python install 3.14Choose a parent directory for the proxy (for example, a ~/Projects folder), then clone the repository, create an .env from the example, and open the directory in your editor:
git clone https://github.com/Pushpenderrathore/claude-code.git
cd claude-code
cp .env.example .env
code .The proxy supports several free upstream providers. NVIDIA NIM is the recommended primary provider, with a free tier of approximately 40 requests per minute — sufficient for sustained coding sessions.
NVIDIA NIM (recommended)
- Create an account at build.nvidia.com and complete phone verification (required for key generation).
- Click Generate API Key, supply a name (for example,
claude-code) and an expiration date. - Paste the key into
NVIDIA_NIM_API_KEYin the Admin UI (or, when running from source, into the corresponding entry in.env).
OpenRouter (optional)
- Create an account at openrouter.ai.
- Generate a key at openrouter.ai/keys and paste it into
OPENROUTER_API_KEY.
Note: Free OpenRouter endpoints can be intermittently rate-limited for accounts without prepaid credit.
DeepSeek (optional)
- Create an account at platform.deepseek.com.
- Generate a key at platform.deepseek.com/api_keys and paste it into
DEEPSEEK_API_KEY.
Save the configuration after entering keys.
Claude Code exposes three model tiers — Opus, Sonnet, and Haiku — each of which may be routed to a distinct provider via MODEL_OPUS, MODEL_SONNET, and MODEL_HAIKU. Any unset tier inherits the value of MODEL.
- Opus (NVIDIA NIM): On build.nvidia.com, filter by Free Endpoint, open the desired model (for example, GLM 4.7), and use the model identifier in the form
nvidia_nim/<model-name>. - Sonnet (OpenRouter): Select a top-context free model from openrouter.ai/collections/free-models and use the identifier in the form
open_router/<model-name>. - Haiku (DeepSeek): Use the DeepSeek Anthropic-compatible endpoint with
deepseek/deepseek-reasoner.
For a full provider reference, see Choose a Provider.
In the proxy directory, start the server:
uv run uvicorn server:app --host 0.0.0.0 --port 8082In a separate terminal, change into the project directory in which Claude Code should operate, then launch the CLI with the proxy environment variables set (see Quick Commands). Claude Code will start against the local proxy; the model picker will reflect the configured tiers.
Claude Code prompts for permission before modifying files by default. For longer automated sessions, Sandbox Mode confines write access to the current project directory while still requiring confirmation for paths outside it:
- Inside Claude Code, run
/sandbox. - Select Sandbox with auto-allow.
Use version control alongside Sandbox Mode so that automated changes can be reviewed and reverted as needed.
.env.exampleenumerates environment key names as a read-only reference for contributors. Use the Admin UI to modify managed proxy settings.- Bugs and feature requests should be reported through Issues.
- Keep changes small and accompanied by focused tests.
- Docker integration pull requests will not be accepted.
- README change pull requests will not be accepted; open an issue instead.
- Run the full check sequence prior to opening a pull request.
- The
except X, Ysyntax is restored in the final release of Python 3.14 (not in the 3.14 alpha). This should be considered before opening pull requests.
Released under the MIT License. See LICENSE for the full text.