OpenAI-Compatible API & Tool Compatibility

Check documented OpenAI-compatible setup paths and keep them separate from unverified streaming, tool-calling, image, and agent behavior.

Documentation reviewed Jul 17, 2026Run lightweight relay check

Integration Compatibility Atlas

Statuses below are evidence labels, not invented endpoint test results.

Tool / AgentCustom
base URL
Custom
model ID
BYOKChat
completions
StreamingTool
calling
ImagesDocs review
date
Documented Documented Documented Documented Conditional Conditional Not verifiedJul 17, 2026
Documented Documented Documented Documented Conditional Conditional Not verifiedJul 17, 2026
Documented Documented Documented Documented Conditional Conditional Not verifiedJul 17, 2026
Documented Documented Documented Documented Conditional Conditional Not verifiedJul 17, 2026
Documented Documented Documented Documented Conditional Conditional Not verifiedJul 17, 2026
Documented Documented Documented Documented Not verified Not verified Not verifiedJul 17, 2026
Documented Documented Documented Documented Conditional Conditional Not verifiedJul 17, 2026
Documentation summaryBased on 7 reviewed integrations
Direct setup 342.9%
Limited setup 457.1%
What the feature labels mean Documented Explicitly shown in public documentation. Conditional Depends on model, request shape or separate endpoint testing. Not verified We found no sufficient public evidence in this review.
Evidence boundary

Documentation review proves a configuration path. It does not prove TokenAir model quality, streaming, tool calling or image behavior in that tool.

Review methodology
LIGHTWEIGHT BROWSER CHECK

Check relay protocol integrity

Run three low-token checks that reveal whether an OpenAI-compatible relay preserves the response, streaming, and tool-call behaviors users rely on.

  • Basic response, usage and returned model metadata
  • SSE event stream and proper [DONE] termination
  • Forced tool call with schema-valid arguments
3 requests · about 100 maximum output tokens · no long-context or quality benchmark
Your API key is not uploadedThree bounded requests run directly from this browser. The key stays in page memory and is cleared after the run. This page does not submit anonymous public evidence. Learn more in Data & Methods.
Explore moreCompare the market Calculate cost