This guide shows how to use Telnyx Inference as a custom OpenAI-compatible inference endpoint for NemoClaw. It does not add a named Telnyx provider to NemoClaw; it uses the provider configuration supported by the current NemoClaw installation.
Telnyx endpoints
Telnyx's OpenAI-compatible base URL is:
https://api.telnyx.com/v2/ai/openai
The model catalog is available at
GET /v2/ai/openai/models.
Chat completions are available at
POST /v2/ai/openai/chat/completions.
Telnyx model IDs are returned by the authenticated model catalog. Use the exact ID for the model you want to run; do not guess or substitute a nearby model.
Get a Telnyx API key
Create or manage a Telnyx API key in the Telnyx Mission Control Portal. Keep the key private and do not commit it to this repository.
Configure NemoClaw
Set the custom-provider variables in the shell where you will run NemoClaw.
Replace <model-id> with an exact model ID returned by the Telnyx catalog.
export NEMOCLAW_PROVIDER=custom
export NEMOCLAW_ENDPOINT_URL=https://api.telnyx.com/v2/ai/openai
export NEMOCLAW_MODEL=<model-id>
export COMPATIBLE_API_KEY=<your-telnyx-api-key>
Then run the normal NemoClaw onboarding flow:
nemoclaw onboard
When NemoClaw asks for the inference provider, select the custom or other OpenAI-compatible endpoint option and use the values above. The exact prompts may vary by NemoClaw release.
Verify the Telnyx route
Before onboarding, verify that the key can discover models and that the chosen model is present:
curl --fail --silent --show-error \
https://api.telnyx.com/v2/ai/openai/models \
-H "Authorization: Bearer $COMPATIBLE_API_KEY"
The response contains the model IDs available to the account. Set
NEMOCLAW_MODEL to one of those exact IDs before onboarding.
For the full Telnyx API reference, see the OpenAI-compatible chat completions documentation.
Notes
- The endpoint requires Bearer authentication.
- The model catalog is account-authenticated and can change as model access changes.
- Telnyx and NemoClaw may support different model capabilities; use the model metadata and the Telnyx API documentation for the selected model.
- This guide does not send or include an API key.