AiGo123
Your AI-Powered Dev Team, Right in Your Editor
Get Cheap API Token → AiGo123.net • Local routing → KKSwitch • OpenRouter | Claude | Codex | DeepSeek
AiGo123 is an AI coding agent for VS Code and JetBrains IDEs. It runs a real VS Code Extension Host inside JetBrains IDEs, so the same agent experience works across both platforms.
AiGo123 is part of an ecosystem: the AiGo123.net model gateway, the KKSwitch local routing desktop app, and the AiGo123 agent itself. Use them together or independently.
What's New in v26.4.126
- Use the newest OpenAI, Anthropic and DeepSeek models in JetBrains IDEs or VS Code — search
AiGo123 in the Plugins/Extensions Marketplace.
- AiGo123.net is compatible with Claude Code and other CLI tools.
- Expose skills as slash commands with fallback execution to make reusable workflows faster to trigger.
🌐 Available languages
- English
- Català
- Deutsch
- Español
- Français
- हिंदी
- Bahasa Indonesia
- Italiano
- 日本語
- 한국어
- Nederlands
- Polski
- Português (BR)
- Русский
- Türkçe
- Tiếng Việt
- 简体中文
- 繁體中文
- ...
Table of Contents
What Can AiGo123 Do For YOU?
- Generate code from natural language descriptions and specs
- Adapt with Modes: Code, Architect, Ask, Debug, and Custom Modes
- Refactor & debug existing code
- Write & update documentation
- Answer questions about your codebase with codebase indexing
- Automate repetitive tasks and run terminal commands
- Utilize MCP Servers, Skills and Slash Commands
- Follow project rules such as
AGENTS.md
Modes
AiGo123 adapts to how you work:
- Code Mode: everyday coding, edits, and file ops
- Architect Mode: plan systems, specs, and migrations
- Ask Mode: fast answers, explanations, and docs
- Debug Mode: trace issues, add logs, isolate root causes
- Custom Modes: build specialized modes for your team or workflow
Learn more: Using Modes
Key Features
| Feature |
Description |
| Agentic coding loop |
Plan, edit, run commands, read output and verify results |
| Codebase indexing |
Semantic codebase_search across your project for accurate answers |
| MCP servers |
Connect external tools such as web search or GitHub, via stdio or SSE |
| Skills & Slash Commands |
Reusable workflows triggered from the chat input |
| Auto-approve |
Optionally auto-approve read, write and command operations |
| Terminal integration |
Run and read terminal output with shell integration and timeouts |
| Checkpoints |
Shadow-git snapshots and diff/restore for every task |
| Multi-provider |
Any OpenAI-compatible endpoint, plus AiGo123.net and KKSwitch |
| Quick fixes |
Editor code actions: Explain, Fix, Improve, Add to Context, New Task |
| Multi-language UI |
Catalog, German, Spanish, French, Hindi, Italian, Japanese, Korean, Polish, Portuguese, Turkish, Vietnamese, Simplified & Traditional Chinese |
The agent works through a well-defined tool set (native typed arguments). Highlights:
| Tool |
Purpose |
read_file |
Read files with slice or indentation-aware semantic extraction |
write_to_file |
Create or fully rewrite a file |
apply_diff / edit / search_replace |
Precise, reviewable edits |
apply_patch |
Apply unified-diff style patches |
execute_command |
Run terminal commands with timeout and cwd control |
read_command_output |
Read truncated output; grep-like search and pagination |
search_files |
Regex search across files with glob filtering |
list_files |
List files and directories, optionally recursive |
codebase_search |
Semantic search over the indexed codebase |
new_task |
Delegate work to a sub-task in a chosen mode |
ask_followup_question |
Ask a clarifying question with suggested answers |
update_todo_list |
Track multi-step progress |
switch_mode |
Switch between Code, Architect, Ask, Debug and custom modes |
use_mcp_tool / access_mcp_resource |
Call MCP servers and read MCP resources |
skill |
Activate a discovered Skill |
run_slash_command |
Run a reusable slash command |
generate_image |
Generate images from a prompt |
attempt_completion |
Present the final result of a task |
Quick Start
- Install the extension from the VS Code Marketplace.
- Open the AiGo123 view in the Secondary Side Bar (or run
AiGo123: New Task from the Command Palette).
- Choose a provider and enter your API key — or connect through AiGo123.net / KKSwitch.
- Describe your task in natural language and let the agent work.
Model Providers
AiGo123 works with any OpenAI-compatible endpoint. Commonly used providers include:
- AiGo123 Cloud — the built-in gateway (no separate key needed when signed in)
- Anthropic (Claude), OpenAI (native), OpenAI-compatible, OpenAI Codex
- Google Gemini, Vertex AI, Amazon Bedrock
- DeepSeek, xAI (Grok), Mistral
- OpenRouter, Requesty, Unbound, LiteLLM, Vercel AI Gateway
- Ollama and LM Studio for fully local models
- Qwen Code and the AiGo123 / Roo provider
KKSwitch — local all-in-one API token routing
KKSwitch is a local, all-in-one API token aggregation, routing and AI Agent integration desktop app. It manages upstream providers, models and local Agent tools in one place:
- Zero configuration — start mainstream AI Agent tools without editing JSON or config files
- Protocol conversion — heuristically detect and convert between OpenAI Chat, OpenAI Responses and Anthropic Messages
- Unified model catalog — aggregate, deduplicate and sort models from enabled providers; switching takes effect immediately without restarting your tools
- Isolated runtime — run multiple Codex, Claude and other clients side by side across different workspaces
- Safe by design — redacted diffs, backups, atomic writes and post-write verification; API keys stay local and are never uploaded
- Local proxy — listens on
127.0.0.1 by default (ports 26001–26100) and exposes an OpenAI-compatible /v1 endpoint; LAN mode is opt-in
When KKSwitch is running, AiGo123 detects it automatically and can route requests through it. Download it at www.kkswitch.com (Windows and macOS). KKSwitch is not bundled and is not required.
AiGo123.net — model & API gateway
AiGo123.net is a next-generation large-model gateway and AI asset management platform. It provides API tokens and access to Claude, Codex, DeepSeek, Gemini, Kimi, Qwen and GLM, with multi-protocol support (OpenAI Chat/Responses, Claude Messages, Google Gemini), smart routing with weighted channels and automatic retry, format conversion, and usage dashboards. It is compatible with Claude Code and other CLI tools.
Checkpoints & Safe Editing
Every task gets its own shadow-git checkpoint history, so you can review and roll back with confidence:
- Automatic checkpoints before file writes, edits, diffs and patches
- Restore to any checkpoint, or preview a restore before applying it
- Diff modes:
since first checkpoint, checkpoint vs next, to current, and full
- Isolated from your real repository — nested-git and environment-variable edge cases are handled safely
- Configurable checkpoint timeout
Codebase Indexing
AiGo123 indexes your project so the agent can answer questions and find relevant code with semantic search (codebase_search):
- Adjust the index size with
aigo123.maximumIndexedFilesForFileSearch (default 10,000)
- Tune embedding throughput with
aigo123.codeIndex.embeddingBatchSize (default 60)
- Supported embedders include OpenAI, Gemini, Ollama, OpenRouter, Vercel AI Gateway and OpenAI-compatible endpoints
MCP, Skills & Slash Commands
- MCP — connect Model Context Protocol servers (stdio or SSE) to give the agent extra tools like web search or GitHub; manage them from the MCP tab
- Skills — reusable, mode-scoped instruction sets (
SKILL.md) discovered from global or project locations
- Slash Commands — save repeatable prompts and workflows and trigger them from the chat input
Rules & Custom Instructions
AGENTS.md — project-level agent rules (toggle with aigo123.useAgentRules)
- Mode-specific rules and custom modes for teams
- Import an existing configuration automatically via
aigo123.autoImportSettingsPath
Configuration Reference
AiGo123 exposes its settings under the aigo123.* namespace:
| Setting |
Description |
aigo123.allowedCommands |
Commands that may be auto-executed when auto-approve is on |
aigo123.deniedCommands |
Command prefixes that are always denied (longest match wins) |
aigo123.commandExecutionTimeout |
Max seconds to wait for a command (0 = no timeout, up to 600) |
aigo123.commandTimeoutAllowlist |
Command prefixes excluded from the timeout |
aigo123.preventCompletionWithOpenTodos |
Block task completion while todos are unfinished |
aigo123.vsCodeLmModelSelector |
Vendor/family for the VS Code Language Model API |
aigo123.customStoragePath |
Custom storage location for tasks and settings |
aigo123.enableCodeActions |
Enable AiGo123 quick fixes in the editor |
aigo123.autoImportSettingsPath |
Auto-import a configuration file on startup |
aigo123.maximumIndexedFilesForFileSearch |
Max files indexed for @ file search (default 10,000) |
aigo123.useAgentRules |
Load AGENTS.md files for agent-specific rules |
aigo123.apiRequestTimeout |
Max seconds to wait for API responses (default 600) |
aigo123.newTaskRequireTodos |
Require todos when creating sub-tasks |
aigo123.startup.revealSidebar |
Auto-reveal the Secondary Side Bar on startup (always / never) |
aigo123.codeIndex.embeddingBatchSize |
Batch size for code-index embeddings (default 60) |
aigo123.debug / aigo123.debugProxy.* |
Debug mode and MITM debug proxy for local troubleshooting |
Open Settings → Extensions → AiGo123 to configure them.
JetBrains IDEs
The same AiGo123 agent is available for JetBrains IDEs (IntelliJ IDEA, WebStorm, PyCharm, GoLand, CLion, PhpStorm, RubyMine, DataGrip, Rider, Android Studio). The JetBrains plugin:
- Runs a real VS Code Extension Host with editor, document, diff, terminal and shell-integration sync
- Automatically prepares a portable Node.js runtime (v20.19.2) — no manual installation required
- Supports multiple agent providers: AiGo123 / Roo Code, Cline, Kilo Code and Costrict
- Downloads VSIX packages from multiple sources with resume and automatic failover
- Detects missing JCEF and shows a diagnostic panel instead of failing to load
Search AiGo123 in the JetBrains Plugins Marketplace to install it.
FAQ
Which IDEs are supported?
VS Code (1.106.0+) and JetBrains IDEs 2022.3+ (IntelliJ IDEA, WebStorm, PyCharm, GoLand, CLion, PhpStorm, RubyMine, DataGrip, Rider, Android Studio).
Do I need an AiGo123.net account?
No. AiGo123 works with any OpenAI-compatible endpoint. AiGo123.net is the recommended ready-to-use gateway, and KKSwitch is an optional local router.
Can I use local models?
Yes. Ollama and LM Studio are supported, along with any OpenAI-compatible endpoint.
How do checkpoints work?
Each task maintains an isolated shadow-git history. You can diff or restore to any checkpoint without touching your real repository.
Are my API keys safe?
Keys are stored locally in your editor's secret storage or your own configuration. KKSwitch keeps keys local and never uploads them.
How do I update?
VS Code updates the extension automatically. For JetBrains, use the IDE's plugin update mechanism.
License
See LICENSE.
Enjoy AiGo123!
Happy Coding!
| |