generated: '2026-08-04' method: searched source: https://llmboost.mangoboost.io/docs/features/openai-api docs: https://llmboost.mangoboost.io/docs/running/lbh name: MangoBoost authentication and access model description: >- MangoBoost has no hosted, credentialed API. Every surface is deployed inside the customer's own infrastructure, so authentication is a deployment concern rather than an API concern. The three credentials that do exist are for pulling assets (Hugging Face, container registry) and for activating the commercial license — not for calling the API. There is no API-key issuance, no OAuth, no OIDC, and no token endpoint anywhere on the MangoBoost surface. surfaces: - api: Mango LLMBoost Inference Server API transport: HTTP/1.1 REST on :8000 (default) security_schemes: [] model: none-by-default detail: >- "Drop the Authorization header for local serving, or put LLMBoost behind your own gateway for auth." Vendor examples set api_key="not-needed". The OpenAI client requires an api_key value, so any non-empty string works; the server does not validate it. recommendation: >- Because the server ships with no authentication, network placement IS the access control. Bind it to a private interface and front it with your own gateway (mTLS, API keys, OAuth) before any non-loopback exposure. url: https://llmboost.mangoboost.io/docs/features/openai-api - api: Mango SDK Device API (libmango) transport: In-process C/C++ library against a local PCIe device security_schemes: [] model: os-privilege detail: >- Access is governed by Linux privileges on the device — mango-ctl and mango-smi examples require sudo, and the container needs /dev/kfd and /dev/dri access for the GPU path. There is no network authentication layer. url: https://sdk.mangoboost.io/docs/api/ - api: Mango OPI Storage Bridge gRPC API transport: gRPC on :50051 (HTTP gateway on :8082) security_schemes: [] model: network-scoped detail: >- The vendor guide calls the bridge over plaintext gRPC with grpc_cli against 127.0.0.1:50051 or the DPU SoC address (192.168.1.2:50051), and notes that "OPI API calls can be made from any server with the same network as the DPU SoC." No TLS, token or per-call credential is documented — the management network is the boundary. url: https://sdk.mangoboost.io/docs/guide/docs_opi/guide credentials: - name: HF_TOKEN kind: bearer-token purpose: Hugging Face token for downloading gated or private model weights. Forwarded into the container. obtained: hf auth login, or export HF_TOKEN=... scope_of_use: model download only — never sent to the LLMBoost API. url: https://llmboost.mangoboost.io/docs/quickstart - name: container registry login kind: registry-credential purpose: Pull the LLMBoost server image. obtained: >- Provisioned by MangoBoost — email contact@mangoboost.io with your Docker Hub username to be invited to the LLMBoost image registry, then `docker login`. url: https://llmboost.mangoboost.io/docs/running/lbh - name: LLMBoost license kind: license-file default_path: $LBH_HOME/license.skm purpose: Activates the licensed LLMBoost optimizations. detail: >- "LLMBoost activates a license on first start (a free trial is provisioned automatically) and then serves locally. Keep the host online for first activation." Imported/validated with `lbh login`. Enterprise seats and offline licensing are handled through mangoboost.io/contact. url: https://llmboost.mangoboost.io/docs/running/configuration not_present: - oauth2 - openIdConnect - mutualTLS - api-key-issuance - scopes-or-permissions-model x-evidence: fetched: '2026-08-04' probes: - url: https://llmboost.mangoboost.io/docs/features/openai-api http_status: 200 - url: https://llmboost.mangoboost.io/docs/running/configuration http_status: 200 - url: https://sdk.mangoboost.io/docs/guide/docs_opi/guide http_status: 200 - url: https://www.mangoboost.io/.well-known/openid-configuration http_status: 404