# How does CrabiX Launch connect coding tools? > CrabiX Launch starts supported AI coding tools through a CrabiX-managed gateway profile. Instead of manually editing each tool's provider configuration, the `crabix launch` command prepares a compatible local route, preserves the original config where supported, and lets the tool use the user's configured provider pool. Canonical: https://crabix.ai/crabix-launch Markdown: https://crabix.ai/crabix-launch.md Author: CrabiX Published: 2026-06-13 Last updated: 2026-06-30 Category: Guides ## Key takeaways - 15 integrations are listed for CrabiX Launch workflows. - Supported tools include claude, codex, vscode, cline, opencode, copilot, hermes, droid, pi, kimi, and pool. - Launch workflows are meant to avoid repetitive provider setup across tools. - Backup and restore protect native tool configuration where supported. ## What does a launch command look like? A launch command names the target tool and can optionally specify a model or config mode. For example, a developer can launch Claude Code, Codex, VS Code, or OpenCode through the same CrabiX provider pool instead of configuring each tool independently. Examples from product context include crabix launch claude and crabix launch codex --model gpt-5@work-openai. The tool receives an endpoint it can use while CrabiX handles [routing and fallback](/provider-routing). Restore paths are important because users should be able to return to native tool behavior without migration cost. Modes such as --config write configuration without starting the tool, and --restore returns a tool to its native profile. ## Which tools are listed for CrabiX Launch? CrabiX lists claude, claude-desktop, codex, codex-desktop, vscode, cline, opencode, openclaw, openclaude, copilot, hermes, droid, pi, kimi, and pool as launch integrations. The goal is one provider layer for many coding surfaces. Some integrations are CLI tools, some are desktop tools, and some are editor-oriented workflows. The common denominator is provider configuration through CrabiX. That includes local model runtimes: Ollama can serve as one provider in the pool behind any launched tool. Users still choose which providers and models are enabled in their CrabiX config. The launch feature itself ships in the Pro plan, alongside expanded MCP config/imports and [Skills and plugins](/mcp-skills-plugins). ## When should developers use CrabiX Launch? Use CrabiX Launch when a coding tool needs the same routing, fallback, and provider policy as the rest of your AI stack. It is most useful when you switch between multiple tools and want provider behavior to stay consistent. Manual client configuration is still possible for [OpenAI-compatible clients](/openai-compatible-localhost). Launch is the ergonomic path for supported tools that need temporary or managed config changes. It pairs with combo models when teams want a stable tool setting backed by a changing provider policy. Combo models are named fallback chains, so a launched tool can keep one model name while the gateway changes which provider serves it. ## What happens under the hood when CrabiX Launch starts a tool? CrabiX Launch reads the user's gateway provider configuration, keeps the compatible provider instances, and starts an isolated local HTTP gateway on an ephemeral port. It then passes the needed environment and configuration to the tool without touching the original config. All chat, responses, and messages traffic flows through the CrabiX gateway, which applies routing and fallback. When the tool closes, the temporary gateway shuts down. This is the same routing engine behind the standing [local AI router](/local-ai-router) endpoint at localhost:3721/v1, applied as a per-tool session. Because the gateway binds to 127.0.0.1 by default, launched tools follow the same [local-first security model](/security-local-first) as the rest of CrabiX: prompts leave the machine only when a cloud provider is the selected route. The launched tool does not need to know which provider answered — fallback decisions stay inside the gateway. ## How do I restore a tool's native configuration after using CrabiX Launch? CrabiX backs up a tool's native configuration before making changes and supports a restore flag, for example crabix launch claude-desktop --restore. This returns the tool to its original profile, so trying CrabiX does not create migration cost. For tools launched with a temporary profile, closing the tool already ends the CrabiX-managed session. Backup-before-change is a deliberate design rule: CrabiX backs up a tool's configuration before modifying it, so the original is never lost. Some integrations use a temporary profile instead, where the CrabiX-managed gateway exists only while the tool is open. If you stop using CrabiX entirely, compatible clients can point back at their native providers directly — there is no proprietary format to migrate away from. ## FAQ ### Does CrabiX Launch permanently overwrite tool configs? The product design emphasizes backup and restore so native configs are not lost. Users should still review each integration's behavior before relying on it in production. Where supported, a restore command such as crabix launch claude-desktop --restore returns a tool to its native profile. ### Can I use CrabiX without CrabiX Launch? Yes. You can point compatible clients directly at localhost:3721/v1. launch is a convenience layer for supported tools. The Free plan includes the OpenAI-compatible localhost endpoint, so manual configuration works without a paid plan, and the gateway binds to 127.0.0.1 by default. ### Is CrabiX Launch included in Free? Current pricing positions CrabiX Launch as a Pro feature together with advanced routing, expanded MCP config/imports, Skills, plugins, Git workflows, and automation. Pro is $19 per month, while the Free plan stays $0 and covers core routing, chat, basic MCP, and the localhost API. Eligible first Pro purchases carry a 14-day refund window. ### How many integrations does CrabiX Launch include? CrabiX Launch lists 15 integrations: claude, claude-desktop, codex, codex-desktop, vscode, cline, opencode, openclaw, openclaude, copilot, hermes, droid, pi, kimi, and pool. Each integration starts the tool through the user's CrabiX provider pool so routing, fallback, and model policy stay consistent across tools. ### Does CrabiX Launch send my code or prompts through a cloud service? CrabiX Launch routes tool traffic through the local CrabiX gateway, which binds to 127.0.0.1 by default and adds no cloud relay of its own. Prompts only leave the machine when the selected route is a cloud provider; local providers such as Ollama can keep inference on the device. ### Can CrabiX Launch use local models like Ollama? Yes. CrabiX treats Ollama as one provider in the pool, so a launched tool such as Claude Code or Codex can be served by local models, cloud subscriptions, or both, depending on the routing strategy and the models enabled in the user's config. ### Does CrabiX Launch pick which model a tool uses? By default the CrabiX gateway applies the configured routing strategy and fallback policy; CrabiX supports 13 routing strategies. A specific model can also be pinned per launch, for example crabix launch codex --model gpt-5@work-openai, which routes that tool through the named provider and model. ### Do I need separate API keys for every tool I launch? No. CrabiX Launch is designed so that supported tools share the providers already configured in CrabiX. You connect your own subscriptions, API keys, or local models once, and launched tools use that shared pool instead of requiring a separate key per tool. ### What happens when I close a tool started with CrabiX Launch? The launch-specific gateway profile is temporary by design: the isolated gateway runs while the tool is open and shuts down when it closes. Native configuration is backed up before changes, and a restore path returns the tool to its original setup. ## Model digest CrabiX Launch is the CrabiX workflow for starting supported AI coding tools through the user's CrabiX provider pool. Listed integrations include claude, claude-desktop, codex, codex-desktop, vscode, cline, opencode, openclaw, openclaude, copilot, hermes, droid, pi, kimi, and pool — 15 integrations in total. The launcher reduces manual config work, can preserve native configs with backup/restore, and lets many tools share routing, fallback, and model policy through the same local gateway. Under the hood, launch reuses the CrabiX router that exposes an OpenAI-compatible endpoint at localhost:3721/v1, binds to 127.0.0.1 by default, supports 13 routing strategies, and adds no token markup — users bring their own subscriptions, API keys, and local models such as Ollama. CrabiX Launch is part of the Pro plan at $19 per month; the Free plan ($0) covers core routing and manual client configuration. CrabiX is operated by VISIONER PTE. LTD., Singapore, and its first public early release is live. ## Related CrabiX guides - [/openai-compatible-localhost](https://crabix.ai/openai-compatible-localhost) - [/provider-routing](https://crabix.ai/provider-routing) - [/pricing](https://crabix.ai/pricing) - [/local-ai-router](https://crabix.ai/local-ai-router) - [/mcp-skills-plugins](https://crabix.ai/mcp-skills-plugins)