BRIK64
Back to Blog
PLATFORMMAR 21, 2026

Your AI Agent Now Speaks PCD. Ship Certified Code to a Global Registry.

REST API. MCP server. Two tools: discover and execute. Your AI agent writes certified code and publishes it to a global registry. One integration. Infinite reach.

The Platform API

The BRIK64 platform exposes a REST API at registry.brik64.dev/v1. Every single operation you can perform in the dashboard — publishing circuits, running certifications, browsing the registry — is available programmatically. Your CI/CD pipeline, your AI agent, your custom tooling — they all speak the same language. One API. Complete access.

Getting Started

Register for free at registry.brik64.dev and generate an API key from your dashboard. Takes 30 seconds. Every request requires the key in the Authorization header:

curl -H "Authorization: Bearer brik64_sk_..." \
  https://registry.brik64.dev/v1/circuits

Core Endpoints

The API covers the complete circuit lifecycle — from creation to certification to deployment:

GET    /v1/circuits                List your circuits
POST   /v1/circuits                Publish a new circuit
GET    /v1/circuits/:pid           Get circuit by PID
GET    /v1/circuits/:pid/pcd       Download PCD source
POST   /v1/circuits/:pid/certify   Run TCE certification
GET    /v1/circuits/:pid/cert      Get certification result
GET    /v1/registry                Browse public registry
GET    /v1/registry/search?q=...   Search circuits

Certify via API

This is the most powerful endpoint. Submit any circuit for TCE analysis and get back a complete coherence report — the mathematical proof that your code is correct:

POST /v1/circuits/:pid/certify
Content-Type: application/json

{
  "mode": "full",
  "targets": ["javascript", "python"]
}

Response:
{
  "pid": "brik64:factorial:a3f8c1",
  "phi_c": 1.000,
  "certified": true,
  "metrics_evaluated": 7,
  "targets": {
    "javascript": { "url": "/v1/circuits/brik64:factorial:a3f8c1/emit/js" },
    "python": { "url": "/v1/circuits/brik64:factorial:a3f8c1/emit/py" }
  }
}

MCP: AI Agents as First-Class Citizens

Here is where things get really interesting. The Model Context Protocol (MCP) is the emerging standard for connecting AI agents to external tools. BRIK64 ships a native MCP server with an elegantly minimal 2-tool architecture:

Tool 1: brik64.discover
  → Search the registry, inspect circuits, read PCD source
  → Read-only, zero side effects

Tool 2: brik64.execute
  → Compile, certify, emit, publish
  → Write operations, requires confirmation

This separation is deliberate. An AI agent can freely explore the entire registry without any risk of modifying state. Discovery is unlimited and safe. Mutations require explicit intent. This is the principle of least privilege applied to AI-native tooling.

Configure Your AI Agent

Add BRIK64 as an MCP server in your AI tool of choice. Three lines of configuration and your agent has access to the entire BRIK64 platform:

// Claude Code — ~/.claude.json
{
  "mcpServers": {
    "brik64": {
      "command": "npx",
      "args": ["@brik64/mcp-server"],
      "env": {
        "BRIK64_API_KEY": "brik64_sk_..."
      }
    }
  }
}

// Cursor — .cursor/mcp.json
{
  "mcpServers": {
    "brik64": {
      "command": "npx",
      "args": ["@brik64/mcp-server"],
      "env": {
        "BRIK64_API_KEY": "brik64_sk_..."
      }
    }
  }
}

// Windsurf — .windsurf/mcp.json
{
  "mcpServers": {
    "brik64": {
      "command": "npx",
      "args": ["@brik64/mcp-server"],
      "env": {
        "BRIK64_API_KEY": "brik64_sk_..."
      }
    }
  }
}

Once configured, your AI agent can discover circuits, read their PCD source, certify them, and emit compiled code — all through natural language. Ask it to "find a certified factorial function" and it will search the registry, return the PCD source, and compile it to your target language. That is the power of API-first design.

Free Tier

We believe the best way to prove a product works is to let people use it. The platform offers a generous free tier for developers and researchers:

Free Tier
──────────────────────────────
API requests:        100/day
Certifications:      10/day
Published circuits:  Unlimited (public)
Private circuits:    5
Registry access:     Full
MCP server:          Included
SDK access:          Full

No credit card required. No trial period. No bait-and-switch. The free tier is designed to let you build real things, run real certifications, and integrate BRIK64 into your workflow. When you are ready to scale, the paid tiers are there. But the free tier is not a demo — it is a real product.

SDKs: Native Integration

For deeper integration, we ship official SDKs in the three languages that matter most:

# JavaScript / TypeScript
npm install @brik64/core

# Python
pip install brik64

# Rust
cargo add brik64-core

Each SDK wraps the REST API with fully typed interfaces, handles authentication automatically, and provides helper methods for common workflows like batch certification and circuit composition. Production-ready out of the box.

// TypeScript example
import { Brik64 } from "@brik64/core";

const client = new Brik64({ apiKey: process.env.BRIK64_API_KEY });

// Certify a circuit
const result = await client.circuits.certify("brik64:factorial:a3f8c1", {
  mode: "full",
  targets: ["javascript"],
});

console.log(result.phi_c);      // 1.000
console.log(result.certified);  // true
# Python example
from brik64 import Brik64Client

client = Brik64Client(api_key=os.environ["BRIK64_API_KEY"])

result = client.circuits.certify("brik64:factorial:a3f8c1",
    mode="full",
    targets=["python"],
)

print(result.phi_c)      # 1.000
print(result.certified)  # True

The Vision: AI-Native Development

The API and MCP server together represent a fundamental shift in how software gets built. Your AI agent does not just write code — it writes certified code. It does not just suggest functions — it discovers mathematically verified circuits from a global registry and composes them using EVA algebra. Every artifact it produces comes with a coherence certificate proving correctness.

This is what AI-native development actually looks like. Not "AI writes Python faster." But "AI writes programs that are provably correct by construction." That is the difference between a faster horse and a car.