generated: '2026-08-05'
method: derived
source: openapi/0g-labs-router-openapi.yml
docs: https://docs.0g.ai
note: >-
  Standards conformance derived from the published OpenAPI plus explicit claims in the 0G docs.
  0G publishes no security certifications (no SOC 2, ISO 27001, PCI DSS, HIPAA or FedRAMP claim
  was found on any 0G host), so no Compliance pointer is emitted — its assurance story is
  hardware attestation (TEE) and on-chain settlement rather than an audited compliance program.
standards:
- id: openapi-3.0
  conforms: true
  evidence: >-
    openapi: 3.0.0 with 21 paths / 23 operations, 126 component schemas, and both security
    schemes defined and applied per-operation. Published at
    https://0gfoundation.github.io/0g-router/openapi.yaml (200, text/yaml, 183,793 bytes).
- id: openai-chat-completions-api
  conforms: true
  evidence: >-
    POST /v1/chat/completions with the OpenAI request/response shape, streaming, tool calling,
    JSON mode (ResponseFormat) and reasoning tokens. GET /v1/models returns the OpenAI list
    envelope. Docs state any OpenAI client works by changing base_url and api_key.
- id: anthropic-messages-api
  conforms: partial
  evidence: >-
    POST /v1/messages with AnthropicRequest/AnthropicResponse, content blocks, usage and cache
    creation. Partial because only models whose supported_formats include the Anthropic format
    are reachable there; others return 400 naming the formats they do serve.
- id: openai-rate-limit-headers
  conforms: true
  evidence: >-
    X-RateLimit-Limit-Requests / -Remaining-Requests / -Reset-Requests on every inference
    response, plus Retry-After on 429.
- id: rfc6750-bearer-token
  conforms: true
  evidence: 'Authorization: Bearer <token> on both securitySchemes (ApiKeyAuth sk-, ManagementKeyAuth mk-).'
- id: oauth2
  conforms: false
  evidence: >-
    No oauth2 securityScheme in the spec; the authentication docs state explicitly "No OAuth
    flow, no wallet signature per request, no session tokens."
- id: oidc
  conforms: false
  evidence: /.well-known/openid-configuration returns 404 on every 0G host.
- id: rfc9457-problem-details
  conforms: false
  evidence: >-
    Errors use the OpenAI error envelope ({error:{message,type,code}, request_id}) with
    application/json, not application/problem+json.
- id: rfc9116-security-txt
  conforms: false
  evidence: /.well-known/security.txt returns 404 on 0g.ai, docs.0g.ai and router-api.0g.ai.
- id: rfc8594-sunset-header
  conforms: false
  evidence: >-
    No Sunset/Deprecation header advertised. Deprecation is announced in prose — the OpenAPI
    info.description Changelog and inline docs warnings.
- id: rfc8615-well-known
  conforms: false
  evidence: No /.well-known/ document is served on any 0G host.
- id: llmstxt
  conforms: partial
  evidence: >-
    /llms.txt is served on both https://0g.ai (1,693 bytes) and https://docs.0g.ai (18,405
    bytes), plus /llms-full.txt (602,979 bytes) on the docs host. Partial because every Table
    of Contents link in the docs llms.txt carries a /docs/ path prefix that 404s — the index is
    unresolvable as published.
- id: a2a-agent-card
  conforms: false
  evidence: >-
    /.well-known/agent-card.json and /.well-known/agent.json return 404 on 0g.ai, docs.0g.ai,
    router-api.0g.ai and pc.0g.ai.
- id: model-context-protocol
  conforms: false
  evidence: >-
    No first-party hosted MCP server was found. 0G ships Agent Skills instead (14 published
    SKILL.md files at github.com/0gfoundation/0g-agent-skills). Third-party 0G MCP servers exist
    but are not 0G-operated.
- id: agent-skills
  conforms: true
  evidence: >-
    14 provider-authored Agent Skills plus an AGENTS.md orchestration guide published at
    https://github.com/0gfoundation/0g-agent-skills, and a second skill package at
    https://github.com/0gfoundation/0g-compute-skills.
- id: eip-1474-ethereum-json-rpc
  conforms: true
  evidence: >-
    https://evmrpc.0g.ai answered eth_chainId with 0x4115 (16661) on 2026-08-05. Docs state
    Hardhat, Foundry, Truffle, ethers.js and viem work unchanged.
- id: erc-20
  conforms: true
  evidence: Wrapped0GBase precompile at 0x…1001 wraps the native token as ERC-20 with quota-based mint and burn.
- id: erc-7857
  conforms: true
  evidence: >-
    0G authors and implements ERC-7857 (encrypted NFT metadata) as the standard behind Agentic
    IDs. https://docs.0g.ai/developer-hub/building-on-0g/agentic-id/erc7857
- id: erc-8004
  conforms: true
  evidence: >-
    "0G officially supports ERC-8004, the Trustless Agent standard. 0G's ERC-8004 registries are
    listed in the official standard repo."
    https://docs.0g.ai/developer-hub/building-on-0g/agentic-id/erc8004
- id: intel-tdx-attestation
  conforms: true
  evidence: >-
    GET /v1/models returns tee_attested: true, tee_type: TDX, tee_verifier: dstack per model,
    and verifiability: TeeML. verify_tee: true makes the Router verify the provider's TEE
    signature synchronously and report tee_verified in x_0g_trace.
- id: op-stack-da
  conforms: true
  evidence: Documented OP Stack rollup integration against 0G DA.
- id: arbitrum-nitro-da
  conforms: true
  evidence: Documented Arbitrum Nitro rollup integration against 0G DA.
- id: eigenlayer-avs
  conforms: true
  evidence: Documented EigenLayer AVS built on 0G DA.
- id: babylon-avs
  conforms: true
  evidence: Documented Babylon AVS built on 0G DA.
certifications_published: []
certifications_note: >-
  No SOC 2, ISO 27001, PCI DSS, HIPAA, FedRAMP, GDPR or CSA STAR claim was found on 0g.ai,
  docs.0g.ai or any 0G subdomain. probe-security-programs.py returned trust=none. No Compliance
  or TrustCenter pointer is emitted.