Setup Guide
OpenCode Go (Zen) API Key on Hermes Agent: One Key, Entire Catalog
OpenCode Go — also called Zen — is an aggregated AI gateway at opencode.ai/zen/go that bundles Kimi, Qwen, GLM, DeepSeek, MiMo, and MiniMax models under a single sk- key. This guide shows you how to get that key and add it on the API Keys page so every model in the Zen catalog is immediately available across your Hermes Agent instances.
What Is OpenCode Go (Zen)?
OpenCode Go (also known as Zen) is a curated model aggregation gateway run by opencode.ai. Rather than managing separate API keys for Moonshot AI, Alibaba Cloud, Zhipu AI, DeepSeek, and MiniMax, you authenticate once with a single sk- Bearer key and gain access to the entire Zen catalog through one consistent OpenAI-compatible endpoint.
The catalog is curated for coding agents and long-context reasoning. Models like DeepSeek V4 Pro (1M context), MiniMax M3 (512K context), and Kimi K2.6 are particularly well-suited for large codebase analysis, multi-file refactoring, and agentic coding workflows running inside Hermes Agent. For a broader comparison, see the AI coding plans compared guide.
What Your OpenCode Go Key Unlocks on Hermes Agent
Once you add your OpenCode Go key on the API Keys page, the following models become available across your Hermes Agent instances:
- Kimi K2.6 (
kimi-k2.6) — Moonshot AI’s flagship model, strong on coding and long-context tasks. - Qwen3.6 Plus (
qwen3.6-plus) — Alibaba Cloud’s top-tier Qwen model for reasoning and instruction following. - MiniMax M2.7 (
minimax-m2.7) — Fast MiniMax chat model via the Zen routing layer. - MiniMax M3 (
minimax-m3) — MiniMax’s flagship with 512K context window, ideal for large codebases. - DeepSeek V4 Pro (
deepseek-v4-pro) — DeepSeek’s most capable model with a 1M token context window. - DeepSeek V4 Flash — Faster, lighter DeepSeek V4 variant for high-throughput tasks.
- GLM-5.1 (
glm-5.1) and GLM-5 — Zhipu AI’s GLM generation, strong at Chinese-English bilingual tasks. - MiMo V2 Omni / Pro / 2.5 — MiMo multimodal reasoning variants for mixed text and image tasks.
- MiniMax M2.5 — Lightweight MiniMax model for fast, cost-efficient completions.
How to Get an OpenCode Go API Key
- Visit opencode.ai/go — Open opencode.ai/go in your browser.
- Create or sign into your account — Register with your email address or OAuth provider.
- Navigate to API Keys — Open the API Keys or credentials section of the dashboard.
- Generate a new key — Click Create Key. Your key is shown once — copy it immediately and store it securely.
- Confirm the prefix — OpenCode Go keys start with
sk-. Verify the copied value before leaving the page.
Add Your OpenCode Go Key on OpenClaw Launch
- Sign in at openclawlaunch.com and open the API Keys page.
- Find the OpenCode Go card. Paste your
sk-key and click Save. We validate the key against the OpenCode Go endpoint live before storing it. - Click “Set as primary model” and pick a model from the Zen catalog (e.g.
kimi-k2.6ordeepseek-v4-pro) to route your running Hermes Agent instances to it immediately.
Your key is encrypted at rest and applied to your existing running Hermes Agent instances — no redeploy needed.
Using OpenClaw instances instead? See the OpenCode Go setup guide for OpenClaw instances.
Troubleshooting
401 Unauthorized / Invalid Key
- Confirm your key starts with
sk-and contains no trailing spaces or newline characters. - Regenerate the key from the opencode.ai dashboard and save the fresh value on the API Keys page.
- Make sure you are using an OpenCode Go key and not a key from a different provider — different
sk-issuers use the same prefix but different signing secrets.
Model Not Found
- Use exact model IDs:
kimi-k2.6,qwen3.6-plus,minimax-m3,deepseek-v4-pro,glm-5.1. Guessed or partial IDs will return a 404. - Confirm a valid OpenCode Go key is saved on the API Keys page first — the models are only routed once the key is confirmed.
Quota or Rate Limit Errors
- Each underlying model in the Zen catalog has its own upstream rate limits. If you receive a 429, wait a few seconds and retry, or switch to a different model in the catalog.
- Check your opencode.ai account dashboard for usage and plan details.
Frequently Asked Questions
Why use OpenCode Go on Hermes Agent instead of individual provider keys?
One key, one billing relationship, one endpoint — instead of separate accounts for Moonshot AI, Alibaba Cloud, Zhipu AI, DeepSeek, and MiniMax. For Hermes Agent coding workflows this is especially convenient: switch between Kimi, DeepSeek, and Qwen without ever leaving the Hermes dashboard or touching the API Keys page again.
What does an OpenCode Go key look like?
OpenCode Go keys start with sk-. They authenticate you against the opencode.ai/zen/go gateway endpoint. OpenClaw Launch automatically uses the correct base URL when you save a key in the OpenCode Go card — you do not need to configure the endpoint manually in Hermes Agent.
Can I use OpenCode Go alongside other provider keys on the same Hermes Agent instance?
Yes. Provider keys are stored independently. You can save an OpenCode Go key alongside Agnes, Groq, OpenRouter, or any other provider. Switch between active models using the “Set as primary model” control or the Hermes Agent instance model selector.
Which OpenCode Go model is best for coding on Hermes Agent?
For large codebases, deepseek-v4-pro (1M context) and minimax-m3 (512K context) are the strongest choices. For speed and daily coding tasks, kimi-k2.6 and qwen3.6-plus are popular picks. See the AI coding plans compared guide for a side-by-side breakdown.
What's Next?
- OpenCode Go on OpenClaw — Same key, set up for OpenClaw instances instead
- AI coding plans compared — Kimi K2.6 vs DeepSeek V4 Pro vs Qwen3.6 Plus and more, side by side
- Hermes Agent BYOK overview — All supported providers and how bring-your-own-key works on Hermes Agent
- API Keys page — Add or manage all your provider keys in one place