Backplane
Backplane Docs https://backplane.gsmlg.net

Backplane Docs

Public gateway references for MCP tools, model traffic, skills, agent setup, and authentication.

Tools/mcp

MCP Hub

Streamable HTTP, OAuth discovery, and namespaced tools.
Models/v1

LLM proxy

Compatible model APIs with resource-bound access tokens.
Google GenAI

Google GenAI and agy

Use the native Gemini Developer API through Backplane, or keep the existing Google Antigravity (agy) subscription route explicit and separate.

Gemini Developer API

The google-gemini-developer provider uses the official Gemini GenerateContent API at /v1beta. Configure an API-key credential in the admin provider settings; Backplane injects it upstream and never forwards caller credentials. Clients may authenticate with Authorization: Bearer, x-api-key, or the Google SDK's x-goog-api-key compatibility header using the same Backplane client token.

Native Google GenerateContent operations

The native Google surface supports model listing, model details, generateContent, streamGenerateContent with alt=sse, and countTokens. Use the Google-shaped request and response bodies; this route does not silently translate to OpenAI.

Google Antigravity (agy)

The google-antigravity provider is a separate OAuth subscription surface for the existing agy workflow. It uses its own native Cloud Code protocol and credential; do not reuse a Gemini API key or treat Antigravity as the public Gemini Developer API.

Configure your client

  1. In the admin LLM provider page, create Google Gemini Developer API with the google-gemini-developer preset, bind an API-key credential, and confirm the base URL ends in /v1beta.
  2. Grant the client llm::models and/or llm::invoke, then use the Google resource path under /v1beta with the Backplane bearer token.
  3. For agy, select the separate google-antigravity preset and its Google OAuth credential. Keep its model aliases and project configuration separate from the Gemini API-key provider.

Request examples

List Google models

curl -H "Authorization: Bearer $LLM_ACCESS_TOKEN" https://backplane.gsmlg.net/v1beta/models

Generate content

curl -X POST \
  -H "Authorization: Bearer $LLM_ACCESS_TOKEN" \
  -H "Content-Type: application/json" \
  https://backplane.gsmlg.net/v1beta/models/gemini-2.5-flash:generateContent \
  -d '{"contents":[{"role":"user","parts":[{"text":"Hello from Backplane"}]}]}'

Stream or count tokens

curl -N -X POST https://backplane.gsmlg.net/v1beta/models/gemini-2.5-flash:streamGenerateContent?alt=sse \
  -H "Authorization: Bearer $LLM_ACCESS_TOKEN" \
  -H "Content-Type: application/json" \
  -d @request.json

curl -X POST https://backplane.gsmlg.net/v1beta/models/gemini-2.5-flash:countTokens \
  -H "Authorization: Bearer $LLM_ACCESS_TOKEN" \
  -H "Content-Type: application/json" \
  -d @request.json

Google GenAI SDK base URL

# Point the SDK at the Backplane origin and pass the Backplane bearer token.
# The native Google path is /v1beta; do not use /v1/chat/completions.
BACKPLANE_GOOGLE_BASE_URL=https://backplane.gsmlg.net
BACKPLANE_LLM_ACCESS_TOKEN=$LLM_ACCESS_TOKEN

API reference

GET /v1beta/models GET /v1beta/models/:model POST /v1beta/models/:model:generateContent POST /v1beta/models/:model:streamGenerateContent?alt=sse POST /v1beta/models/:model:countTokens