Models & protocols
OpenAI Chat / Responses, Anthropic Messages, and Gemini / Vertex. Public model names with supported conversion or native forwarding.
Model accessENTERPRISE AI GATEWAY · SELF-HOSTED
Give enterprise AI
room to grow.
One gateway for models, access, cost, and security. Give platform teams a shared view and business teams the tools to put AI to work.
Go gateway / Bilingual console / Docker Compose
YOUR AI INFRASTRUCTURE● Unified control
Business appsAI agentsInternal tools
XXHubEnterprise AI Gateway
OpenAIAnthropicGeminiMore providers
PostgreSQLRedisSelf-hosted
✓Multiple protocols✓Organization governance✓Programmable guardrails✓Agents & multimodal
BUILT FOR YOUR ORGANIZATION
One endpoint. A broader enterprise toolkit.
Connect model resources, people, policies, and request outcomes for the everyday work of managing enterprise AI.
Governance & autonomy
Organizations, teams, projects, and scoped roles. Personal and service virtual keys with model permissions, expiration, and access revocation.
PermissionsCosts with clear ownership
Usage by user, key, team, and project. Budgets, RPM / TPM, cache rates, time windows, and historical price snapshots.
Pricing & usageRouting for each business
Weight, load, latency, cost, and usage policies. Inherited templates, model groups, categorized fallbacks, retries, and cooldowns.
Routing policiesPolicies in the request path
Text rules, secret detection, and moderation adapters. Combine HTTP, JSON, and LLM checks in XGo to block, redact, or flag.
Programmable guardrailsContinuous agent calls
Multi-turn messages and tool-call structures, Responses history continuation, and session tracing. Validate streams and compare models in Playground.
Sessions & runtimeBeyond conversation
Deployment-specific images, audio, embeddings, reranking, and video tasks. Track asynchronous status, results, usage, and settlement.
Media tasksVisibility for operations
Model catalogs, configuration diagnostics, request and error logs, cost details, audit records, and health checks.
OperationsTHE XHUB DIFFERENCE
Distinct capabilities. Practical outcomes.
Policies you can program
Compose enterprise checks in XGo pre-call guardrails. Draft testing and real calls share the same executor.
Explore scripts & examples →Routing along your organization
Select templates through key, team, and organization scopes. Clear a binding to inherit; routing never expands model permissions.
Understand inheritance →Pricing evidence for every call
Keep applied price snapshots, distinguish response and prompt caches, and deduplicate asynchronous task settlement.
Understand pricing →ONE WORKSPACE
Shared resources. Scoped responsibilities.
Administrators configure resources. Members discover available models. Use one console to connect, validate, and trace requests.
Actual interface from a configured instance. Fresh installations need provider credentials and deployments.
FROM SETUP TO SCALE
A clear path to adoption.
Connect models
Add credentials and deployments. Publish model names.
Set boundaries
Configure teams, members, access, and budgets.
Integrate apps
Issue virtual keys. Validate in Playground.
Keep governing
Review costs, adjust policies, and trace failures.
From one application to your organization.
Enterprise AI platform
Share a model catalog with departmental budgets and clear resource ownership.
Assistants & coding agents
Manage service keys, tool-call structures, and multi-turn usage.
RAG & content production
Bring retrieval, dialogue, and media tasks into a shared endpoint and usage system.
START ON YOUR INFRASTRUCTURE
Start in your own environment.
Clone the repository, build images on your host, then start the gateway, console, and stores with Docker Compose.
Requires Git, Go 1.25, Node.js ≥24.14.1, npm ≥11.10.0, Docker Compose, Bash, and a readable system CA bundle (override with XHUB_CA_CERT_FILE). Use WSL2 on Windows.
Complete setup & first request →Console
:3000Gateway API :4000Docker Compose
git clone https://github.com/qiniu/xhub.git
cd xhub
bash deploy/build.sh
docker compose up -d --no-buildPython · OpenAI SDK
from openai import OpenAI
client = OpenAI(
api_key="YOUR_XHUB_VIRTUAL_KEY",
base_url="http://localhost:4000",
)
response = client.chat.completions.create(
model="YOUR_PUBLIC_MODEL_NAME",
messages=[{"role": "user", "content": "Hello, XHub!"}],
)
print(response.choices[0].message.content)DOCUMENTATION
Understand the platform. Use its capabilities.
Setup, administration, protocols, permissions, security, and operations. Built directly from repository documentation.
Quickstart
Prerequisites, Docker, source installation, and your first call.
Administration
Models, organizations, teams, keys, routing, and costs.
API & integration
Registered endpoints, protocol capabilities, and request paths.
Developer guide
Protocol registration, request lifecycle, source locations, and tests.