Documentation

Point any OpenAI SDK at https://api.pgsgrove.com/v1 and go. Start with the quickstart.

Getting started

• Quickstart: Point any OpenAI SDK at the API and send your first request.
• Coding tools & SDKs: Wire up opencode, OpenClaw, Hermes, and any OpenAI-compatible tool.
• Native protocols: Anthropic Messages, OpenAI Responses, and Ollama native on the same key.

Making requests

• Request parameters: What passes through untouched, and the few fields the proxy sets.
• Streaming: Receive tokens as they are generated over SSE.
• Structured output: JSON mode and function calling.
• Reasoning models: How thinking is returned: the reasoning_content field.
• Multimodal input: Send images to vision-capable models.

Models & pricing

• Models & pricing: The GET /v1/models endpoint, the model id reference, and the price table.
• Turbo: Priority routing for lower, steadier latency.
• Prompt caching: Automatic prefix caching at a reduced input rate.
• Speech & embeddings: Text to speech (kokoro-82m) and 768-dim embeddings.
• How usage is measured: Refill, bank, and pacing mechanics, plus per-model relative draw rates on the coding plan.

Operations

• Rate limits: Per-account request and concurrency limits, and how to handle them.
• Errors & retries: Status codes, 402/429 handling, and backoff.

Legal

• Terms of Use: The Open Grove API Terms of Use.