Integrations

Integrations overview

Any agent that can talk to an OpenAI-compatible endpoint can be governed by Spineforge: you only change a base URL and a key.

The universal rule

Whatever you are connecting (an IDE, a command-line tool, a script, a framework or a workload running in the cloud) the integration is the same two lines:

1

Base URL: point your agent at the Spineforge gateway

https://spineforge-backend.onrender.com/gateway/v1
2

API key: use your Spineforge gateway key

sfgw-xxxx-xxxx

Select OpenAI (or OpenAI Compatible) as the provider, set the Base URL to the gateway URL, use your gateway key as the API key, and name your model. The request then flows:

Your agent / IDE / CLI  ──►  Spineforge gateway  ──►  LLM provider
   (gateway key)             authenticate · scope check ·        (your real key, resolved
                             budget & rate limits · log           from the Vault)

Choose your environment

IDEs & Editors Cursor, VS Code native chat, VS Code extensions such as Continue, and other OpenAI-compatible editors.CLI Tools OpenCode, Antigravity CLI, and any CLI that reads OPENAI_BASE_URL / OPENAI_API_KEY style settings.Custom Agents OpenAI SDKs in Python and Node.js, LangChain, raw HTTP, and visual builders such as Flowise and Langflow.Cloud-Hosted Agents Docker, Kubernetes, serverless and CI. Set the base URL and key as environment variables, no code change.Model Naming How to name models so the gateway routes to the right provider.Gateway Keys & Vault Create, scope, budget and revoke the keys you hand to agents.

Before you integrate anything

  1. Add your real provider key (OpenAI, Anthropic, Google Gemini, Groq…) to the Dashboard Vault.
  2. Create a gateway key (sfgw-...) in the dashboard's API Keys tab, ideally one per agent or environment.
  3. Pick a model name. Names containing a / need a provider prefix. See Model Naming.

What you get for free

  • Every request attributed to an agent, with provider, model, tokens, latency and cost
  • Per-agent budget caps and rate limits enforced inline
  • Provider and model scopes per key
  • Instant revocation: one click blocks every in-flight and future request
  • Streaming responses passed straight through
Heads up: credential custody only holds if your raw provider keys live exclusively in the Spineforge Vault. If a key also lives in an env file or shared secret store, any process that can read it can bypass the gateway. Remove it from those locations once agents are migrated.

Not on the list?

If your tool lets you set a custom OpenAI-compatible base URL and API key, it will work. If it doesn't, or you are unsure, talk to us and we'll help you connect it.