One API key, one OpenAI-compatible endpoint, and provider-specific model IDs for Cerebras, Qwen, Kimi, OpenAI, GLM, NVIDIA, and future routes. Use these docs to wire Jatevo into Hermes Agent, OpenCode, Cline, Roo Code, OpenClaw, LiteLLM, and normal SDKs.
Every Jatevo route uses the same base URL and Bearer token shape. Choose the model ID that maps to your provider.
Works with any API Master key
export JATEVO_API_KEY=sk-jvo-your-key-here
curl https://api.jatevo.ai/v1/chat/completions \
-H "Authorization: Bearer $JATEVO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "cerebras/gemma-4-31b",
"messages": [
{"role": "user", "content": "Say hello from Jatevo API Master."}
],
"max_tokens": 120
}'Use the standard /v1 path for SDKs and harnesses. Provider routing is controlled by the model ID.
OpenAI-compatible chat endpoint for most SDKs, agents, and coding tools.
Responses-style endpoint where the selected upstream route supports it.
List models visible to the current API key.
Machine-readable route examples and model metadata.
Each provider needs its own model IDs, limits, pricing, and caveats. These cards are the first version of that catalog.
Cerebras
Use the 131K context setting. Jatevo validation found that to be the practical limit on the current Cerebras route.
Alibaba Cloud
Use the Jatevo Qwen route instead of putting Alibaba keys in local agent configs.
Jatevo Inference
Best default for coding agents that need long-context source edits and tool-use style prompts.
OpenAI via Jatevo LB
Use this route for frontier reasoning and OpenAI-style compatibility while keeping keys and budgets centralized.
Z.ai / GLM
Keep this as an explicit route. It is no longer the OpenClaw default model.
NVIDIA route
Use when you need a large reasoning route and can tolerate provider-specific reasoning controls.
Each guide explains where to paste the base URL, where the key belongs, which model ID to use, and how to verify the connection.
Best first model: cerebras/gemma-4-31b
custom_providers:
- name: jatevo
base_url: https://api.jatevo.ai/v1
key_env: JATEVO_API_KEY
api_mode: chat_completions
models:
cerebras/gemma-4-31b:
name: Cerebras Gemma 4 31B
context_length: 131000
model:
provider: custom:jatevo
default: cerebras/gemma-4-31b
context_length: 131000Best first model: kimi-k2.7-code
{
"$schema": "https://opencode.ai/config.json",
"provider": {
"jatevo": {
"npm": "@ai-sdk/openai-compatible",
"name": "Jatevo API Master",
"options": {
"baseURL": "https://api.jatevo.ai/v1"
},
"models": {
"kimi-k2.7-code": {
"name": "Kimi K2.7 Code"
},
"cerebras/gemma-4-31b": {
"name": "Cerebras Gemma 4 31B"
}
}
}
}
}Best first model: cerebras/gemma-4-31b
API Provider: OpenAI Compatible
Base URL: https://api.jatevo.ai/v1
API Key: sk-jvo-your-key-here
Model ID: cerebras/gemma-4-31b
Max output: 32768
Context window: 131000Best first model: openai/gpt-5.5 or provider-prefixed routes
{
"models": {
"providers": {
"jatevo-api": {
"api": "openai-completions",
"baseUrl": "https://api.jatevo.ai/v1",
"apiKey": "${JATEVO_API_KEY}",
"models": {
"cerebras/gemma-4-31b": {},
"kimi-k2.7-code": {},
"qwen/qwen3.7-max": {}
}
}
}
}
}Best first model: Any Jatevo model ID
model_list:
- model_name: jatevo-gemma
litellm_params:
model: openai/cerebras/gemma-4-31b
api_base: https://api.jatevo.ai/v1
api_key: os.environ/JATEVO_API_KEYBest first model: Any Jatevo model ID
from openai import OpenAI
import os
client = OpenAI(
base_url="https://api.jatevo.ai/v1",
api_key=os.environ["JATEVO_API_KEY"],
)
response = client.chat.completions.create(
model="cerebras/gemma-4-31b",
messages=[{"role": "user", "content": "Reply with one short sentence."}],
max_tokens=120,
)
print(response.choices[0].message.content)Use this as the index for future dedicated provider and harness pages.
| Provider | Harness | Status | Model ID | Context | Streaming |
|---|---|---|---|---|---|
| Cerebras Gemma 4 31B | Hermes Agent | Ready | cerebras/gemma-4-31b | 131K | Yes |
| Cerebras Gemma 4 31B | Cline / Roo Code | Ready | cerebras/gemma-4-31b | 131K | Yes |
| Kimi K2.7 Code | OpenCode | Ready | kimi-k2.7-code | 256K listed | Yes |
| Qwen 3.7 Max | OpenClaw | Ready | qwen/qwen3.7-max | 1M listed | Yes |
| OpenAI / Codex LB | OpenClaw | Ready | lb/gpt-5.5 | 272K | Yes |
| GLM route | LiteLLM | Draft | glm/GLM-5.2 | 200K listed | Yes |
Users paste their Jatevo API Master key into the usage dashboard to inspect remaining credit, request logs, cached input tokens, and per-model usage. Vendor keys stay server-side.
These are the first checks when a harness fails to connect.
Set Authorization: Bearer sk-jvo-... and verify the key is active.
Top up or wait for the budget window to reset.
Use the model ID enforced on the key, especially for grant keys.
Reduce agent concurrency or wait for the rolling RPM window.
Keep prompt plus completion under the provider context limit.
Retry. Jatevo logs the upstream route and request ID for debugging.
The current /docs page is the index. These should become dedicated pages as the docs surface grows.