SCP

The problem

Every LLM-controlled body today is welded to one environment. Change the body and you rebuild everything: the prompt, the tool schema, the cache, the plumbing. There was no open protocol for it.

The insight

Let the body run at 60Hz. Push events up only when it cannot answer locally. The brain teaches. The muscle remembers.

Familiar situations are handled locally. Novel situations wake the brain. Cost is proportional to novelty.

Install

npm install scp-protocol          # one body
npm install @srk0102/plexa        # many bodies under one brain

Quick start (5 minutes)

const { SCPBody, PatternStore } = require("scp-protocol")
const { OllamaBridge } = require("scp-protocol/bridges/ollama")

class Patrol extends SCPBody {
  static bodyName = "patrol"
  static tools = {
    advance: { description: "move forward", parameters: { speed: { type: "number", min: 0, max: 1, required: true } } },
    halt:    { description: "stop", parameters: {} },
  }
  async advance({ speed }) { console.log(`advancing at ${speed}`) }
  async halt()             { console.log("halted") }
}

const body = new Patrol({
  patternStore: new PatternStore({ featureExtractor: (e) => ({ kind: e.kind }) }),
})

// Same situation twice. Second time is cached.
const entity = { kind: "obstacle" }
let hit = body.decideLocally(entity)
if (!hit) {
  const bridge = new OllamaBridge({ model: "llama3.2" })
  // body.learnFromBrain(entity, "halt")   // pretend LLM said halt
  body.learnFromBrain(entity, "halt")
  await body.invokeTool("halt")
} else {
  await body.invokeTool(hit.decision)
}

// Next tick, same entity:
hit = body.decideLocally(entity)
console.log(hit)   // { decision: "halt", confidence: 0.05, source: "exact" }

Expected output:

halted
{ decision: 'halt', confidence: 0.05, source: 'exact' }

When to use what

Plexa is built on scp-protocol. Anything an SCP body does inside Plexa works identically outside.

Not just robotics

SCP works for any system that runs continuously and pushes events:

Game NPCs   Robot arms   Web servers   Log monitors   API gateways   Any loop

If it ticks and emits events, SCP connects it to any LLM. See examples/web-backend in the docs.

Adapters tested

Five adapters, two languages (JS and Python), one protocol.

Docs

Full documentation: https://srk-e37e8aa3.mintlify.app

Pages cover the three-layer architecture, the pattern store and adaptive memory, bridges, the adapter contract, the full API, and four complete walkthroughs (cart-pole, two bodies, web backend, and more).

License

MIT