WEB6 API Reference

Completions

{
  "provider": "auto",
  "model": "auto",
  "messages": [
    {"role": "system", "content": "You are a helpful OASIS guide."},
    {"role": "user", "content": "What is holonic memory?"}
  ]
}

data: {"content":"Once ","done":false}
data: {"content":"upon ","done":false}
data: {"content":"a time...","done":false}
data: {"content":"","done":true,"promptTokens":12,"completionTokens":47}

Embeddings

Image Generation

Video Generation

Audio

Search

Reranking

Moderation

Translation

Classification

Structured Extraction

Document Processing

Code Execution

Batch Processing

Memory

Guardrails

Fine-Tuning

GraphRAG

Prompt Optimisation

FAHRN — Reasoning Network

Model Catalogue

{
  "total": 23,
  "models": [
    {
      "id": "gpt-4o",
      "name": "GPT-4o",
      "providerName": "OpenAI",
      "minPlanLabel": "Bronze",
      "inputPer1kUSD": 0.0025,
      "outputPer1kUSD": 0.01,
      "contextWindow": 128000,
      "supportsVision": true,
      "supportsFunctions": true
    }
  ]
}

Provider Status

Cost Estimate

{ "estimatedUSD": 0.001250, "note": "Based on gpt-4o pricing at $0.0025/1K in + $0.01/1K out" }

Usage

{
  "dailyCallsUsed": 47,
  "dailyCallLimit": 1500,
  "remainingCallsToday": 1453,
  "karmaMultiplier": 3.0,
  "monthlySpendUSD": 0.042,
  "dailyTokensUsed": 185000,
  "periodMonth": "2026-08",
  "periodDay": "2026-08-23"
}

Health

{
  "status": "ok",
  "version": "2.0",
  "timestamp": "2026-08-23T10:00:00.0000000Z"
}

OpenServ

[
  {
    "id": "oasis-web6/gpt-4o",
    "name": "GPT-4o via WEB6",
    "provider": "openai",
    "contextWindow": 128000,
    "capabilities": ["chat", "tools", "vision"]
  },
  ...
]

Orchestrators

Route multi-agent orchestration frameworks through WEB6. Each endpoint translates a framework-native payload into WEB6 completions and returns results in the originating framework's expected shape — no framework SDK required.

{
  "process": "sequential",
  "tasks": [
    {
      "id": "research",
      "agent": "Researcher",
      "description": "Find the top 3 OASIS providers for video generation",
      "expected_output": "A ranked list with reasons"
    },
    {
      "id": "write",
      "agent": "Writer",
      "description": "Turn the research findings into a blog section",
      "expected_output": "200-word blog paragraph"
    }
  ]
}

{
  "isError": false,
  "result": {
    "state": {
      "research_output": "...",
      "__node__": "research",
      "__timestamp__": "2026-08-23T10:00:00Z"
    },
    "nextNode": "summarise"
  }
}

MCP — Model Context Protocol

WEB6 exposes its full capability surface as an MCP server, making every tool callable directly from Claude Code, claude.ai connectors, OpenAI Responses API, and any other MCP-compatible client — with zero HTTP configuration required.

Connect in Claude Code:

claude mcp add oasis-web6 \
  --transport http \
  --url https://api.web6.oasisomniverse.one/mcp \
  --header "Authorization: Bearer <your-jwt>"

Or in ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "oasis-web6": {
      "transport": "http",
      "url": "https://api.web6.oasisomniverse.one/mcp",
      "headers": { "Authorization": "Bearer <your-jwt>" }
    }
  }
}

MCP Tool Reference

All 35+ tools exposed by the MCP server. Each tool is callable from any MCP client — Claude Code, claude.ai, OpenAI, or your own agent.

Rate Limits & Headers

Every response includes rate-limit context headers:

Quota exceeded returns HTTP 429 with Retry-After seconds until midnight UTC.

All responses use the OASISResult envelope:

{
  "isError": false,
  "message": null,
  "result": { /* endpoint-specific data */ }
}

Webhook Delivery & HMAC Signing

Async endpoints (video generation, batch jobs, fine-tuning) accept a callbackUrl field. When the job completes, WEB6 POSTs a signed WebhookPayload to that URL:

{
  "jobId":          "abc123",
  "jobType":        "video",
  "status":         "completed",
  "result":         { /* job-specific result */ },
  "error":          null,
  "completedAtUtc": "2026-08-23T10:05:22Z"
}

When the WEB6_WEBHOOK_SECRET environment variable is set on the server, every delivery also includes:

Verify the signature in your callback endpoint:

# Python verification example
import hmac, hashlib

def verify_webhook(body: bytes, signature: str, secret: str) -> bool:
    expected = "sha256=" + hmac.new(
        secret.encode(), body, hashlib.sha256
    ).hexdigest()
    return hmac.compare_digest(expected, signature)