Official developer gateway connecting sovereign LLM inference, ephemeral JIT prompt synthesis, 52+ specialized MCP tools, and multimodal perception across 330+ global edge nodes.
Every core subsystem is exposed through unified prefixes under https://api.floworkos.com.
Deploy lightweight worker nodes to secondary Cloudflare accounts. Bypass subrequest limits, distribute heavy AI crawls, and keep execution 100% locked to your Master User ID.
A pure, lightweight edge worker script (~80 lines) ready to deploy to any Cloudflare account with built-in JIT Memory pulling and Trajectory auto-sealing.
// swarm/cf-worker/_worker.js
export default {
async fetch(request, env) {
const MASTER_USER_ID = (env.MASTER_USER_ID || '<YOUR_USER_ID>').toLowerCase();
const BRAIN_SSOT = env.BRAIN_SSOT_URL || 'https://api.floworkos.com';
const authHeader = request.headers.get('Authorization') || '';
// Strict User Lock Guard
const body = await request.json().catch(() => ({}));
if (body.userId && body.userId !== MASTER_USER_ID) {
return new Response('Forbidden: Node locked to @' + MASTER_USER_ID, { status: 403 });
}
// Pull 6-Layer Ephemeral JIT Memory from Brain
const jit = await fetch(BRAIN_SSOT + '/v1/brain/flowork-jit', {
method: 'POST',
headers: { 'Authorization': authHeader, 'x-user-id': MASTER_USER_ID },
body: JSON.stringify({ prompt: body.prompt, userId: MASTER_USER_ID })
}).then(r => r.json()).catch(() => ({}));
// Execute via Neural Router and Seal to Trajectory Ledger
const res = await fetch(BRAIN_SSOT + '/v1/chat/completions', {
method: 'POST',
headers: { 'Authorization': authHeader, 'x-user-id': MASTER_USER_ID },
body: JSON.stringify({ model: 'coder', messages: [{ role: 'user', content: body.prompt }] })
}).then(r => r.json());
return new Response(JSON.stringify({ success: true, result: res.choices[0].message.content }));
}
};
Configure your target Cloudflare account with your master user ID. The worker automatically registers itself into your Swarm Registry.
# swarm/cf-worker/wrangler.toml name = "flowork-swarm-node-1" main = "_worker.js" compatibility_date = "2024-09-23" [vars] BRAIN_SSOT_URL = "https://api.floworkos.com" MASTER_USER_ID = "<YOUR_USER_ID>"
Follow hands-on walkthroughs to integrate LLM inference, MCP tools, and sensory perception into your stack.
Learn how to authenticate with your personal key and stream OpenAI-compatible chat completions with zero setup.
Configure Claude Desktop or Cursor IDE to stream 52+ specialized MCP tools via live Server-Sent Events (SSE).
Extract compact semantic markdown from heavy PDFs, Word docs, CSV tables, and images before sending to LLMs.
Synthesize 6-layer cognitive prompts and retrieve 5W1H user memories dynamically without bloating token windows.
Embed passwordless Flowork SSO into your web or mobile app using our lightweight client JavaScript SDK.
Query active model capabilities, context limits, and fallback rotators across the global sovereign mesh.
Built upon biological separation of concerns: Brain, Nervous Router, MCP Tools Fleet, Sensory Perception, and Identity.
Ephemerality-first cognitive synapse handling 6-layer prompt synthesis, 5W1H long-term memories, and trajectory checkpoints.
Sub-20ms multi-provider LLM inference engine with monotonic tool call indexing and resilient fallback rotators.
52+ specialized tools, modular playbooks, and GitOps registry delivering autonomous filesystem, web, and search capabilities.
Multimodal sensory processor extracting structured markdown from heavy PDFs, Word docs, CSV tables, UI images, and audio.
Decentralized identity gateway with instant ready-to-paste SDKs for web and mobile client applications.
World-class pair-programming interface with Monaco Code Studio, real-time Canvas, and multi-model chat.