Skip to main content
Docsdeveloper
Canonical developer setupInstaller-first workflow

Developer docs now start from the install center

Use /dev-tools as the canonical setup guide for local AI coding tools. That page replaces scattered manual instructions for Claude Code, Codex, OpenClaw, and OpenCode, explains optional MCP, skills, agent-team, and opt-in auto-update, and documents how model selection now includes mainstream vendor leaders alongside DeepAILab brand products.

Open Install CenterReview Model Catalog
One-command install center
Canonical

Canonical setup for Claude Code, Codex, OpenClaw, and OpenCode. Start here before editing local config files manually.

Open
API quickstart
Start here

Get an API key, make your first request, and move from install to the first successful response quickly.

Open
Model catalog
Models

Compare mainstream vendor flagships and DeepAILab brand models, then choose your own default model strategy.

Open
API Center
Keys & usage

Manage API keys, inspect usage, and persist your product-side default model preferences.

Open

What changed

The developer docs are now organized around one consistent provisioning workflow.

Manual edits to ~/.claude, standalone MCP scripts, and separate OpenCode sync flows are now compatibility paths, not the primary documentation path.

Users can choose whether to install MCP, skills, agent-team defaults, hooks, and auto-update instead of inheriting everything by default.

Runtime support still differs by client. OpenClaw keeps MCP and hooks manual in phase 1, while Claude Code, Codex, and OpenCode have broader managed pack coverage.

Model defaults now surface OpenAI, Anthropic, Google, DeepSeek, and DeepAILab product models side by side so users can decide what fits their workload.

The canonical model commands are maintained in /dev-tools and should be treated as the source of truth.

Self-hosted teams that run the frontend in multiple instances should configure RATE_LIMIT_REDIS_URL or REDIS_URL so proxy rate limiting remains authoritative across nodes.