Provider Support
This page summarizes the current public provider surface in one reference table so you can scan coverage and caveats together.
Read it conservatively:
- provider support is route-dependent
- all five providers support sync and streaming text generation
- advanced operations (image generation, audio, embeddings) vary by provider — OpenAI has the broadest coverage
- proxy routing controls are not universal across providers
- proxy streaming is exposed for all five public proxy providers
- being listed here does not mean every provider feature is exposed publicly
- image edits, multipart image uploads, and Responses API image-tool support are not part of the public OpenAI image surface
This table keeps customer-facing capability and limitation notes while omitting adapter internals.
| Provider | Proxy Path | Executions | Proxy Routing | Proxy Streaming | Limitations |
|---|---|---|---|---|---|
| Anthropic | /v1/proxy/anthropic | ✓ | ✓ | ✓ | Public proxy support is text messages with sync and streaming behavior. Public cross-provider fallback is limited to the implemented OpenAI and Anthropic text subset. |
| elevenlabs | /v1/executions | ✓ | ✕ | ✕ | ElevenLabs Direct is available on /v1/executions for synchronous TTS/STT only. Hosted ElevenLabs Agents and realtime sidecar governance are out of scope for this connector. |
/v1/proxy/google | ✓ | ✕ | ✓ | The proxy route currently exposes text generation with sync and streaming behavior plus an image-understanding subset. /v1/executions additionally supports embed.text and transcribe.audio. | |
| Meta | /v1/proxy/meta | ✓ | ✕ | ✓ | Current public proxy support is text generation with sync and streaming behavior. Non-text parity remains narrower and route-dependent. metadata.adapter_mode selects native, compatible, or byo. |
| OpenAI | /v1/proxy/openai | ✓ | ✓ | ✓ | Public image support is the same narrow OpenAI image-generation subset on /v1/execute, /v1/executions, and /v1/proxy/openai: generation only, explicit non-'auto' size and quality, with unsupported pricing shapes denied. Image edits remain fail-closed; multipart uploads and Responses API image-tool support remains out of scope. Also the only current public streaming path on /v1/executions. |
| vocal_bridge | /v1/executions | ✓ | ✕ | ✕ | Vocal Bridge Direct governs outbound voice calls on /v1/executions and advertises realtime sessions plus the public Vocal Bridge platform feature inventory in the capability registry. Administrative agent, log, recording, evaluation, and debug APIs are represented as connector features rather than callable execution operations. |
| xAI | /v1/proxy/xai | ✓ | ✕ | ✓ | Current public support is text or chat with sync and streaming behavior. |
Execution operations
- Anthropic
- generate.text
- elevenlabs
- generate.audio, transcribe.audio
- embed.text, generate.text, transcribe.audio, understand.image
- Meta
- embed.text, generate.text, understand.image
- OpenAI
- embed.text, generate.audio, generate.image, generate.text, transcribe.audio, understand.image
- vocal_bridge
- call.outbound
- xAI
- generate.text
Cross-check route-specific behavior on the authored pages when you need request examples, headers, or current operational caveats.
Last updated on Edit this page on GitHub