Your agents are going to do something wrong. Novyx Core makes it recoverable.
Novyx Core is the control plane for coding agents — one package that makes them persistent, inspectable, and safe to let loose.
Tamper-evident audit chain, policy-gated actions, transactional rollback. Works with whatever runtime you're already using.
Get started in 30 seconds
uvx novyx-mcpRead the quickstart →MCP memory works locally with zero config. No API key needed.
Governed actions · Novyx Core
Works with
Four things it does
Act Safely
Write governance rules in YAML. Approve high-risk actions before they fire. Cryptographic audit for everything.
- ✓Policy-as-code (custom YAML rules)
- ✓Approval workflows (Solo/Team/Enterprise)
- ✓SHA-256 audit chain
Recover
Point-in-time rollback. Preview before executing. Like git for agent state.
- ✓Restore to any point in time
- ✓Replay time-travel
- ✓Eval health scoring
Coordinate
Context Spaces let agents share memory. Agent A learns it, Agent B knows it.
- ✓Shared memory spaces
- ✓120 MCP tools
- ✓Works in Claude Code & Cursor
Remember
Persistent memory with semantic search. Survives sessions, restarts, crashes.
- ✓Sub-100ms recall
- ✓Semantic search
- ✓Knowledge graph
See it in action
Save what your agent learns (memories), corrupt them on purpose, restore the clean state in one call (rollback), then check the tamper-evident log that proves nothing slipped through (cryptographic proof) — all in your browser. No signup required.
Try Interactive Demo →Get started in 30 seconds
Setup: Install & initialize
pip install novyx
from novyx import Novyx
nx = Novyx(api_key="nram_...")Then 3 lines: request, approve, inspect
result = nx.request_action(
"github.issue.create", payload={...}
)
nx.approve(result)
nx.action_audit(result["id"])Your agents act. You stay in control.
Free tier. No credit card. Start in 30 seconds.
5,000 memories free. Forever.
Join our Discord