Claude Code local routing dashboard

clco-helper

The local dashboard that lets Claude Code use any AI model. Pick a provider, choose a model route, and launch — all from the terminal.

Get started Source

clco dashboard · 96x28

clco v0.3.82 Ready

⠋

Claude Code → clco:8080 → Codex / GPT-5.5

proxy active / session 3 / AI services 2 / accounts 4

RUN — MODELS — TOOLS — STATUS — SETTINGS — EXIT

Get started

Quick connect

[ 0 ]Codex OAuth (GPT-5.5)easiest · GitHub subscription

[ 1 ]DeepSeek API keyeasy · V4 Pro recommended

[ 2 ]OpenAI API keyeasy · GPT-5.5, GPT-4.1

[ 3 ]Google Gemini APIeasy · Gemini 3.1 Pro

Routing

[ 4 ]Route Mesh multi-modelauto-spread · rate-limit aware

[ 5 ]OpenRouter / Poe / Vercel300+ models via routing API

[ 6 ]Local model (Ollama/LM Studio)offline · local GPU

Tier → Provider

opus→deepseek / V4 Pro

sonnet→codex / GPT-5.5

haiku→gemini / 3.1 Flash

●Route Mesh active

● request.complete│opus → DeepSeek V4 Pro│437ms

/dashboard

Claude Code localhost:8080 clco-helper Provider API

Fast start

Install and run

Install then run clco to pick your provider, choose a model route, and launch Claude Code in one flow.

curl -fsSL \
  https://raw.githubusercontent.com/claudianus/clco-helper/main/scripts/install.sh | sh
clco

bun install -g clco-helper
clco

powershell -NoProfile -ExecutionPolicy Bypass `
  -Command "irm .../install.ps1 | iex"
clco

bunx --bun clco-helper

npx clco-helper

What it does

One proxy for every Claude Code route

Passthrough Proxy

Anthropic-compatible providers keep Claude Code request/response format intact. Zero translation overhead — streaming, tool use, system prompts, and thinking blocks pass through natively. Works with Anthropic API clones.

OpenAI Translator

Translates between Anthropic Messages API and OpenAI Chat Completions API in real time for 15+ providers. Converts tool definitions, streaming chunks, thinking tokens, and usage counts bidirectionally.

Route Mesh

Intelligent traffic spreader that monitors rate limits, timeouts, and concurrency across multiple models and providers. Auto-shifts traffic when providers throttle — your session stays alive under pressure.

Tier Mapping

Map Claude Code's Opus, Sonnet, Haiku tiers to any provider and model combination. Each tier independently configurable with per-provider API keys, rate limits, and retry policies.

Live Dashboard

Real-time TUI dashboard: active routes, tier mappings, token usage, latency, error rates, and provider health. Inspect headers, retries, and cost per session — color-coded for instant clarity.

Web Search MCP

Add a free, zero-config web search MCP server to Claude Code from the MCP screen. DuckDuckGo, Brave Search, and SearXNG backends — search, fetch, and cite sources via MCP protocol.

Multi-account Vault

Securely store multiple API keys per provider with per-account quota tracking and automatic rotation on rate limit. Encrypted at rest on your machine.

Claude Settings Manager

View, edit, backup, and restore Claude Code settings from the TUI. Auto-backup before changes, one-click restore — no manual JSON editing.

Deep Research Packs

Pre-configured MCP server packs for arXiv, PubMed, Semantic Scholar, and web crawling — install from the MCP screen with auto-detection of Python/uvx/node.

mattpocock/skills

One-click install the mattpocock/skills collection to boost Claude Code capabilities. From TOOLS → Install Skills in the TUI — zero config, fully automatic.

plug & play tools

MCP Server Packs — one toggle, full capability

Pre-configured MCP servers that auto-detect your environment. No config files, no manual setup — just toggle on from the TUI.

◎

Deep Research arXiv + PubMed + Semantic Scholar + web crawling. Full academic research pipeline in one pack. bunx clco-deep-research-mcp

■

Productivity Deep research + sequential thinking + knowledge graph + filesystem. The complete thinking toolkit. 4-in-1 bundle

▶

Brave Search Web + image + video + news search via Brave Search API. Free tier included. API key required

◆

GitHub Search repos, code, issues, PRs. Create/manage repositories, files, and branches via MCP. GitHub token

◯

Firecrawl Web scraping and crawling with AI-powered extraction. Convert any website to structured markdown. API key required

▷

Playwright Headless browser automation. Click, type, screenshot, and interact with any web page via MCP. Chromium auto-install

◇

Knowledge Graph Persistent memory for Claude Code. Store entities, relations, and observations across sessions. local SQLite

◈

Sequential Thinking Structured reasoning with revision tracking. Break complex problems into iterative thought chains. zero-config

quick answers

Frequently asked questions

Does clco-helper modify Claude Code?

No. clco-helper does not patch, modify, or wrap Claude Code binaries. It uses documented settings and a local proxy daemon.

Which AI providers are supported?

30+ providers across three integration modes: Anthropic-compatible, OpenAI-compatible, and native adapters (Codex OAuth, Gemini, Vertex, Bedrock).

Can I use multiple AI accounts?

Yes. Store multiple API keys per provider with usage tracking and automatic rotation on rate limit. Keys are encrypted at rest.

Does it work with Claude Code updates?

Yes. clco-helper has a built-in Claude Code update checker and self-updater in the RUN menu. Checks npm with one-hour cache.

Is there telemetry or data collection?

None. clco-helper runs entirely on your machine. No analytics, no telemetry, no usage data sent anywhere. Outbound connections only go to your configured AI providers.

Windows / Linux / macOS?

All three. macOS/Linux via curl or npm/bun. Windows via PowerShell install script or npm. WSL, Git Bash, and native PowerShell all supported.

Supported paths

30+ providers, three integration modes

Anthropic-compatible Anthropic, DeepSeek Anthropic, MiniMax Anthropic passthrough

OpenAI-compatible OpenAI, OpenRouter, Poe, Vercel, Groq, Together translate

Native adapters Codex OAuth, Gemini, Vertex, Bedrock provider-specific

Full provider list

AnthropicOpenAICodex OAuthDeepSeekGeminiGoogle Vertex MiniMaxKimiGLM/Z.AIQwenOpenRouterPoe Vercel AI GatewayCloudflare AI GatewayGroqxAITogetherFireworks MistralHugging FaceNVIDIAPerplexityCohere Azure FoundryAWS BedrockAlibaba Coding PlanTencent Coding Plan Z.AI Coding PlanXiaomi Token PlanSiliconFlowOllamaLM Studio

Featured providers

🅰 Anthropic Claude Opus 4.7 · Sonnet 4.6 · Haiku 4.5

⭕ OpenAI GPT-5.5 · GPT-5.5 Pro · GPT-4.1

⬡ Gemini Gemini 3.1 Pro · 3 Pro · 2.5 Flash

⬩ DeepSeek DeepSeek V4 Pro · V4 Flash · R1

⟐ Codex OAuth GPT-5.5 · GPT-5.5 Pro via Copilot

◇ OpenRouter 300+ models · unified API gateway

Local first

Claude Code unchanged, routes under your control

clco-helper does not patch Claude Code binaries. It uses documented ~/.claude/settings.json settings and a local daemon to manage provider/model routes. Writes backups before any change and can restore from the menu.

READY Claude Code → clco:8080 → Codex / GPT-5.5

request.complete · 0 errors · statusLine synced