All three GPT-5.6 variants are now available, with updated pricing and context-tier rules View pricing
HopBase
Log in
Production AI Infrastructure

One gatewayfor Claude, Codex, Seedance, and Gemini

Run coding agents alongside API, image, and video workloads through one reliable access layer. Centralize routing, keys, quotas, failover, and usage across every team.

hopbase — quickstart
# Keep Claude Code. Change one environment variable.
$ export ANTHROPIC_BASE_URL="https://api.hop-base.com/v1"
$ export ANTHROPIC_AUTH_TOKEN="sk-hb-xxxxxxxxxxxxxxxxxxxxxxxxxxxx"
$ claude "Refactor dependency injection in this module and add unit tests"
routeclaude-sonnet-5Claude Max (primary; automatically fails over to AWS Bedrock)
policyOrganization · Project key · Team quota enterprise-group / proj-key-a3f / quota-ok
usage1,248 input · 386 output tokens · logged (attributed to project and team)
Read 18 files across 3 packages
Mapped four dependency boundaries
Updated constructor bindings and test doubles
Added success and failure-path coverage
24 tests passed in 1.8s
Done. The refactor is ready for review
API endpoint https://api.hop-base.com/v1
One endpoint. Keep your SDKs and workflows.
OpenAI SDKAnthropic SDKClaude CodeCodex
Built for production

Models evolve. Your infrastructure stays stable.

HopBase provides one stable control layer for routing, failover, access controls, and usage tracking across coding, API, image, and video workloads.

01

Scale without routing bottlenecks

Handle batch workloads, shared development environments, and traffic spikes across business lines. Automatic multi-channel failover maintains availability, with concurrency sized to your workload.

02

Fail over without changing code

Use stable, continuously maintained channels for Claude, Codex, OpenAI, and AWS Bedrock, built for long-running workloads and advanced reasoning.

03

Govern every team and key

Segment API keys, quotas, and billing by organization, project, team, or customer. Use the same controls for internal platforms and multi-tenant products.

04

Track every request and token

Track every call by model usage for departmental chargebacks, customer billing, margin analysis, and usage audits.

05

Keep the SDKs you already use

Keep your OpenAI or Anthropic SDK and change only the base URL to route through one access layer.

Pricing

Choose your model. Compare channels.

You pay the discounted channel rate × actual usage. Official list prices are shown only as a benchmark. Because rates vary by channel, review one channel at a time.

Model ID Official list price Max limited-time offer Savings vs. list Notes
claude-fable-5 Flagship$10 / $50$3 / $16Read $0.31 · Write $3.92 / $6.26save ~69%Latest high-reasoning model
claude-opus-5 New$5 / $25$1.5 / $8Read $0.16 · Write $1.96 / $3.13save ~69%Next-gen Opus workhorse
claude-opus-4-8 Recommended$5 / $25$1.5 / $8Read $0.16 · Write $1.96 / $3.13save ~69%Top reasoning for heavy Claude Code tasks
claude-opus-4-7$5 / $25$1.5 / $8Read $0.16 · Write $1.96 / $3.13save ~69%1M-context Opus
claude-opus-4-6$5 / $25$1.5 / $8Read $0.16 · Write $1.96 / $3.13save ~69%1M-context Opus
claude-opus-4-5-20251101$5 / $25$1.5 / $8Read $0.16 · Write $1.96 / $3.13save ~69%Short alias claude-opus-4-5
claude-sonnet-5 New$3 / $15$0.95 / $4.5Read $0.094 · Write $1.18 / $1.88save ~69%Next-gen Sonnet workhorse
claude-sonnet-4-6 Balanced$3 / $15$0.95 / $4.5Read $0.094 · Write $1.18 / $1.88save ~69%Balanced workhorse for long context
claude-sonnet-4-5-20250929$3 / $15$0.95 / $4.5Read $0.094 · Write $1.18 / $1.88save ~69%Short alias claude-sonnet-4-5
claude-sonnet-4-20250514 Deprecated / auto-routed to 4.6$3 / $15$0.95 / $4.5Read $0.094 · Write $1.18 / $1.88save ~69%Legacy ID auto-routes to claude-sonnet-4-6
claude-haiku-4-5-20251001$1 / $5$0.3 / $1.5Read $0.031 · Write $0.39 / $0.63save ~69%Lightweight, low-cost model

› Prices are in USD per 1 million tokens (input / output). Model availability is shown in the console.

llms.txt

One prompt. Any coding agent. A working setup.

Give this prompt to Claude Code, Codex, Cursor, or another coding agent. It will read the HopBase llms.txt index, find the relevant docs, and answer with the right model and endpoint.

hopbase — llms.txt
Please read HopBase's documentation index below and answer my questions based on it: https://hop-base.com/llms.txt About HopBase: an enterprise AI gateway for Claude, Codex / OpenAI, Gemini, GLM, Seedream image and Seedance video models — one OpenAI- and Anthropic-compatible BASE_URL with unified keys, group governance and usage-based billing. My question: how do I connect Claude Code through HopBase?
FAQ

What teams ask before they switch

Straight answers on scale, reliability, pricing, and integration. See the docs for implementation details, or contact us about your workload.

Yes. HopBase provides one gateway for enterprises, one-person companies (OPCs), and teams across multiple business lines. Gateway-level groups, quotas, and cost tracking support batch jobs, shared team access, and production workloads. Concurrency is sized to your use case.

Most API proxies simply route requests through a single account, which limits reliability and operational control. HopBase adds high-concurrency routing, automatic multi-channel failover, scoped keys, quotas, and cost tracking for sustained production use.

HopBase keeps personal accounts out of your integration. You connect with enterprise API keys; AWS Bedrock requires no personal Claude login, and other Claude, Codex, and OpenAI channels are governed by group permissions.

HopBase supports Claude Fable 5, Opus 5, Sonnet 5, and the Opus and Haiku families; OpenAI GPT and Codex models; Gemini; GLM 5.2; Seedream image models; and Seedance video models. AWS Bedrock currently supports Opus models only. See the console for the complete model list.

Claude Max offers the lowest limited-time rate at about 31% of list for frequent, cost-sensitive workloads. AWS Bedrock connects directly to AWS without a personal login, costs about 51% of list, and currently supports Opus models only. Claude Max (full quota) / Kiro provides higher quotas at about 37% of list. Choose based on budget, model coverage, and reliability requirements.

HopBase rates are approximately 31% of list for Claude Max, 51% for AWS Bedrock, 37% for Claude Max (full quota) / Kiro, 6.6% for Codex, and 55% for GLM 5.2. Each percentage is your rate divided by the comparable official list price, using ¥6.8/USD where conversion is required. Billing is based on actual usage.

No. Prepaid credits have no expiration date and remain available while your account is active.

HopBase provides one integration and one governance layer across providers. Keep Claude Code, Codex, and your OpenAI or Anthropic SDK while adding scoped keys, quotas, cost allocation, and usage auditing.

Start with your workload

Bring your AI stack under one gateway

Tell us what you run, how much concurrency you need, and which models matter. We’ll recommend the right channels and help configure your keys and SDKs.

View pricing
Support QQ 2278168210
Telegram @Aaqhop Open
QQ Community HopBase Community Join

Include "Enterprise access" or "AWS Bedrock" in your message so we can recommend the right group and channel faster.