Keep OpenHands pointed at one OpenAI-compatible address. TEKIZ.AI
chooses among your local models and connected cloud providers based on
what the task needs and what is available.
OpenHands can use
OPENAI_BASE_URL=https://api.tekiz.ai/v1 with a
hosted generated sk_tekiz_* key, or
OPENAI_BASE_URL=http://127.0.0.1:8790/v1 for local
routing.
Local Ollama fallback
Use local-first for low-risk background work,
balanced for general agent sessions, and stronger
authorized cloud routes for fragile repo-wide changes.
Keep working when an account is busy
When an OpenHands route, provider key, or upstream model rate-limits
or fails, TEKIZ.AI can retry the next healthy eligible provider
while OpenHands stays on one endpoint.
Set up OpenHands setup
export OPENAI_BASE_URL=https://api.tekiz.ai/v1
export OPENAI_API_KEY=sk_tekiz_your_generated_key
# Then, in OpenHands Settings > LLM, use:
# Provider: OpenAI
# Model: tekizai/frontier
# Base URL: https://api.tekiz.ai/v1
# API key: your generated sk_tekiz key
# Use an OpenHands secret/environment setting for a generated key.
Development-mode OpenHands configuration
Development-mode OpenHands can instead use this TOML in its
config.toml. It is a file example only. Do not paste it
into a terminal.
OpenHands can continue through provider outages because TEKIZ.AI
handles retries and backup routes behind the same address.
Use more than one account key
Authorized credential pools can load balance and fail over on rate
limits, provider health, latency, and policy.
Use accounts you are allowed to use
TEKIZ.AI does not grant unauthorized OpenAI, Anthropic, Gemini,
NVIDIA, Ollama Cloud, or other provider access. Customers bring
authorized credentials or use hosted routing for TEKIZ.AI account
control.
Change routed models from one OpenHands setup
OpenHands stays pointed at one address while TEKIZ.AI chooses a
model, prefers local or cloud routes as requested, and switches when a
provider fails.