Strix Custom LLM Setup

Route Strix through TokenHub with environment variables and confirm native tool calling on an authorized target.

Create TokenHub API Key

What is Strix?

Strix is an AI security testing agent that coordinates multi-step, tool-driven application security assessments.

This Strix setup uses the official LiteLLM-style STRIX_LLM value with a TokenHub API base and key. It requires structured native tool_calls, and should only be used for authorized security testing. Strix coordinates security tools through an agent loop, which makes structured native tool calls a hard compatibility requirement. Run it only against systems you own or are explicitly authorized to assess, keep credentials private, and verify the selected scope and model behavior with a controlled first test.

Strix uses LiteLLM to reach more than 100 providers. For an OpenAI-compatible API, it does not provide a generic provider settings page. Configure STRIX_LLM, LLM_API_KEY, and LLM_API_BASE in the terminal that starts Strix.

Confirm your authorization to test

Strix is a security-testing agent. Use it only against targets you own or are explicitly authorized to test. Switching its model to TokenHub does not alter that authorization boundary.

Follow Strix's official Custom Base URL setup

export STRIX_LLM="openai/YOUR_TOKENHUB_MODEL_ID"
export LLM_API_BASE="__API_BASE_URL__/v1"
export LLM_API_KEY="sk-..."

strix --target ./your-app

These variables map directly to the official OpenAI Custom Base URL example:

  • STRIX_LLM uses LiteLLM's provider/model-name form. openai/ is the routing prefix; the remainder is the TokenHub model ID.
  • LLM_API_KEY is your TokenHub API key.
  • LLM_API_BASE is the OpenAI-compatible API root including /v1.

Launch and verify from the same terminal

Environment variables apply only to the current shell and its child processes, so start Strix from the terminal where you set them. Begin with a low-risk check against an authorized local target, then confirm the model, endpoint, and usage in TokenHub logs.

If Strix cannot initialize the model, inspect the openai/ prefix in STRIX_LLM. For a 401, inspect LLM_API_KEY. If no request reaches TokenHub, confirm that the process inherited the variables. If the agent generates text but takes no action, choose a model with reliable structured tool calls.

Official reference

Choose a compatible Strix model

WorkloadModel choiceWhy it fits
Complex coding tasksA tool-capable reasoning modelUse a model with reliable tool use, streaming, and the context length your workflow needs.
Daily implementationA balanced coding modelPrioritize reliable edits and useful latency for normal coding loops.
Background and subagent workA fast, lower-cost modelUse this for iterations, summaries, and simpler work only after validating tool use.
Browse TokenHub models and compatibility

Strix LLM provider FAQ

What should STRIX_LLM contain?

Use openai/ followed by the exact TokenHub model ID, for example openai/YOUR_TOKENHUB_MODEL_ID.

Why does the Strix agent make no progress?

The model or inference server may be returning tool calls as plain text. Strix needs structured native tool_calls.

Can I save the values instead of exporting them every time?

Yes. Strix supports configuration through ~/.strix/cli-config.json; use its documented env format and keep the file private.

How do I configure a custom OpenAI-compatible API in Strix?

Set LLM_API_BASE to the TokenHub /v1 root, LLM_API_KEY to the TokenHub key, and start Strix from the environment containing those values.

What authorization is required for Strix testing?

Only assess systems you own or are explicitly authorized to test. Connecting TokenHub does not grant permission to test third-party targets.

References

This setup is based on TokenHub and official documentation, last verified on 2026-09-08.