Quickstart

Start with the product path Novyx is built around now: put a gate in front of one action that can change production.

1. Install

Python

pip install novyx

TypeScript

npm install novyx

curl

No installation needed. Use any HTTP client.

2. Get an API key

Sign up at novyxlabs.com and create an API key.

MCP local mode

If you are using the MCP Server for local development, you can start without an API key. Local mode uses SQLite at ~/.novyx/local.db. Use cloud mode when you need shared audit evidence, approvals, or production action governance.

3. Submit a protected action

Python

from novyx import Novyx

nx = Novyx(api_key="nram_your_key")

result = nx.submit_action(
    "github.merge_pr",
    {
        "repo": "acme/api",
        "pr_number": 42,
        "environment": "production",
    },
    agent_id="deploy-agent",
)

print(result["status"])   # allowed, blocked, or pending_review
print(result["message"])  # reviewer-friendly explanation

TypeScript

import { Novyx } from "novyx";

const nx = new Novyx({ apiKey: "nram_your_key" });

const result = await nx.submitAction(
  "github.merge_pr",
  {
    repo: "acme/api",
    pr_number: 42,
    environment: "production",
  },
  { agent_id: "deploy-agent" },
);

console.log(result.status);  // allowed, blocked, or pending_review
console.log(result.message); // reviewer-friendly explanation

curl

curl -X POST https://novyx-ram-api.fly.dev/v1/actions \
  -H "Authorization: Bearer nram_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "action": "github.merge_pr",
    "params": {
      "repo": "acme/api",
      "pr_number": 42,
      "environment": "production"
    },
    "agent_id": "deploy-agent"
  }'

The response tells your agent whether it may proceed, must wait for approval, or is blocked.

What Novyx records

Evidence	Why it matters
Requested action	The exact operation the agent wanted to run
Policy result	Why the action was allowed, blocked, or sent to review
Approval state	Who approved or denied the action and when
Execution result	What happened after the gate cleared
Recovery context	Checkpoints, rollback notes, or compensation steps when available

Supporting features

Memory, semantic search, replay, eval, and rollback are supporting surfaces for investigation and recovery. Use them when you need context around an action. Do not treat them as a substitute for gating production-changing work before it runs.

Next steps

• Approval workflows — Route risky actions to reviewers
• Custom policies — Define action governance rules
• Actions API — Full request and response fields
• MCP Server — Expose governed actions to MCP-capable agents