MangoBoost

aid: mangoboost
name: MangoBoost
description: 'MangoBoost, Inc. builds full-stack Data Processing Unit (DPU) hardware and AI-infrastructure software for data
  centers. Founded in 2022 and headquartered in Seattle with operations in Canada and Korea, the company ships the Mango BoostX
  DPU family (RoCE AI / RDMA, NVMe-oF initiator and target, TCP Offload Engine), the Mango SDK — a C/C++ device library (libmango),
  CLI tools (mango-ctl, mango-smi) and an OPI-compliant gRPC storage bridge — and Mango LLMBoost, an enterprise LLM inference
  server that exposes a drop-in OpenAI-compatible REST API on customer-owned AMD Instinct GPUs. Its developer surface is self-hosted
  rather than SaaS: the LLMBoost server is distributed as a container plus the llmboost_hub (lbh) Python CLI, and the DPU
  is programmed through libmango and OPI gRPC on the card.'
url: https://raw.githubusercontent.com/api-evangelist/mangoboost/refs/heads/main/apis.yml
image: https://cdn.sanity.io/images/hx87iaks/production/044e3e87d4c32dc3c01996e772f4b3f7c08b873e-376x61.png
x-type: company
x-source: harvest:secondary-market
x-harvest-url: https://forgeglobal.com/mangoboost_stock/
x-tier: profiled
x-tier-reason: enrichment
specificationVersion: '0.20'
created: '2026-08-04'
modified: '2026-08-04'
tags:
- Company
- Artificial Intelligence
- Machine Learning
- Infrastructure
- Data Center
- Semiconductors
- Hardware
- Storage
- Networking
- Inference
- GPU
- DPU
apis:
- aid: mangoboost:llmboost-inference
  name: Mango LLMBoost Inference Server API
  description: 'LLMBoost is MangoBoost''s enterprise LLM inference server. It serves the OpenAI REST API on /v1 so an existing
    OpenAI client migrates with a base-URL change: POST /v1/chat/completions, POST /v1/completions, POST /v1/embeddings, POST
    /v1/responses, POST /v1/audio/transcriptions and /translations, GET /v1/models, plus GET /health and GET /metrics (Prometheus).
    Streaming, JSON-schema structured output, tool/function calling and multimodal image input are supported. The server is
    self-hosted on the customer''s own AMD Instinct GPUs — there is no MangoBoost-hosted endpoint — and is started with `lbh
    serve <Repo/Model>` or `llmboost serve <model>` inside the LLMBoost container.'
  humanURL: https://llmboost.mangoboost.io/docs/features/openai-api
  baseURL: http://localhost:8000/v1
  x-deployment: self-hosted
  x-default-port: 8000
  tags:
  - Inference
  - Artificial Intelligence
  - LLM
  - OpenAI Compatible
  - Self Hosted
  properties:
  - type: Documentation
    url: https://llmboost.mangoboost.io/docs/
  - type: GettingStarted
    url: https://llmboost.mangoboost.io/docs/quickstart
  - type: APIReference
    url: https://llmboost.mangoboost.io/docs/features/openai-api
- aid: mangoboost:mango-sdk
  name: Mango SDK Device API (libmango)
  description: The Mango SDK development package (libmango-dev) provides the C/C++ user API, libraries and example applications
    for programming MangoBoost DPU devices. Documented API groups are Common, Device Context, Memory, DMA Transfer, TCP/IP
    Offload, TLS Offload, POSIX (compatibility shims), AI Image Preprocess, GPU StorageBoost (GSB), RDMA PCC and Software-Defined
    Device. Applications include libmango.h and link with pkg-config; packages are distributed as .deb/.rpm (mango-sdk, mango-cli,
    libmango-dev, mango-drivers, mango-drivers-spdk) obtained by contacting MangoBoost.
  humanURL: https://sdk.mangoboost.io/docs/api/
  x-deployment: on-device-library
  tags:
  - SDK
  - DPU
  - Storage
  - Networking
  - RDMA
  properties:
  - type: Documentation
    url: https://sdk.mangoboost.io/docs/guide/
  - type: APIReference
    url: https://sdk.mangoboost.io/docs/api/
- aid: mangoboost:opi-storage-bridge
  name: Mango OPI Storage Bridge gRPC API
  description: MangoBoost ships an Open Programmable Infrastructure (OPI) gRPC bridge that translates standard OPI storage
    API calls into MangoBoost SDK / SPDK RPC calls on the DPU, so NVMe-oF initiator (NTI) and target (NTT) offloads can be
    configured with the vendor-neutral OPI interface. The bridge listens on gRPC port 50051 with an HTTP gateway on 8082;
    documented calls include CreateNvmeSubsystem, CreateNvmeController, CreateNvmeRemoteController, CreateNvmePath and CreateNvmeNamespace.
    The implementation is open source under the Linux Foundation OPI project as opi-mangoboost-bridge.
  humanURL: https://sdk.mangoboost.io/docs/guide/docs_opi/overview
  x-deployment: on-device-grpc
  x-default-port: 50051
  tags:
  - gRPC
  - Storage
  - NVMe-oF
  - OPI
  - DPU
  properties:
  - type: Documentation
    url: https://sdk.mangoboost.io/docs/guide/docs_opi/guide
  - type: SourceCode
    url: https://github.com/opiproject/opi-mangoboost-bridge
maintainers:
- FN: Kin Lane
  email: kin@apievangelist.com
- FN: APIs.json
  email: info@apis.io
common:
- type: IssueTracker
  url: https://github.com/opiproject/opi-mangoboost-bridge/issues
- type: Releases
  url: https://github.com/opiproject/opi-mangoboost-bridge/releases
- type: License
  name: Apache-2.0
  url: https://github.com/opiproject/opi-mangoboost-bridge/blob/main/LICENSE
- type: DomainSecurity
  url: security/mangoboost-domain-security.yml
- type: Packages
  url: packages/mangoboost-packages.yml
- type: SDKs
  url: packages/mangoboost-packages.yml
- type: CLI
  url: cli/mangoboost-cli.yml
- type: Authentication
  url: authentication/mangoboost-authentication.yml
- type: Conventions
  url: conventions/mangoboost-conventions.yml
- type: ErrorCatalog
  url: errors/mangoboost-error-codes.yml
- type: Lifecycle
  url: lifecycle/mangoboost-lifecycle.yml
- type: Conformance
  url: conformance/mangoboost-conformance.yml
- type: LLMsTxt
  url: llms/mangoboost-llms.txt
- type: Website
  url: https://www.mangoboost.io/
- type: DeveloperPortal
  url: https://sdk.mangoboost.io/
- type: Documentation
  url: https://llmboost.mangoboost.io/docs/
- type: APIReference
  url: https://sdk.mangoboost.io/docs/api/
- type: GettingStarted
  url: https://llmboost.mangoboost.io/docs/quickstart
- type: Support
  url: https://www.mangoboost.io/contact
- type: Blog
  url: https://www.mangoboost.io/resources/blog
- type: GitHubOrganization
  url: https://github.com/MangoBoost
- type: TermsOfService
  url: https://www.mangoboost.io/terms-of-service
- type: PrivacyPolicy
  url: https://www.mangoboost.io/privacy-policy
- type: Careers
  url: https://www.mangoboost.io/careers
- type: News
  url: https://www.mangoboost.io/media/news
- type: Publications
  url: https://www.mangoboost.io/resources/publications
- type: LinkedIn
  url: https://www.linkedin.com/company/mangoboost/
- type: Twitter
  url: https://x.com/mangoboost_inc
x-enrichment:
  date: '2026-08-04'
  status: enriched
  artifacts_added: 10
  pass: local-v1
x-coverage:
  state: covered
  reason: no-machine-readable-spec
  detail: Fully profiled from three live, server-rendered documentation hosts, but MangoBoost publishes no machine-readable
    contract for any of its three API surfaces — /openapi.json, /openapi.yaml, /swagger.json, /api-docs and /redoc all return
    404 on www., sdk. and llmboost.mangoboost.io, and the OPI gRPC bridge ships no first-party .proto. The LLMBoost server
    is OpenAI-compatible, so an OpenAPI describing its /v1 surface would be cheap for them to publish.
  evidence:
  - url: https://llmboost.mangoboost.io/docs/features/openai-api
    status: 200
  - url: https://sdk.mangoboost.io/docs/api/
    status: 200
  - url: https://llmboost.mangoboost.io/openapi.json
    status: 404
  - url: https://www.mangoboost.io/llms.txt
    status: 404
  checked: '2026-08-04'