R-Secure · Self-hosted security for AI coding tools

Let agents build. Secrets stay inside.

  • Self-hosted·On your machines
  • Secrets·Never stored
  • 61·Detection rules
  • Linux·macOS·Windows

R-Secure monitors AI tool activity, blocks unauthorized access to sensitive credentials, and stores all records on your own servers.

Book a demo What leaves your network
Hooks into
  • Claude Code
  • Cursor
  • Google Antigravity
Other AI tools: partial, through the Linux kernel watcher
The leak nobody logs

Your developers' AI tools open .env files, SSH keys and cloud credentials on their own. Nobody is keeping the log.

Example: one Claude Code session, 41 minutes, 212 actions. Three of them touched a secret.

00:0010:0020:0030:0041:12
Inside the network edge

Your activity data never leaves your network.

R-Secure runs on your own machines. Each developer's agent reports to an R-Secure server you host, and that is where every record stays.

Stays inside

Code, prompts and activity

Every event goes from the developer's machine to your R-Secure server, inside your network. Nothing is sent to RAKSYON.

Never stored

Secret values

Files are scanned in memory on the developer's machine and thrown away. When a key or password turns up, only the rule name and an 8-character fingerprint are kept.

Counts only

The one outbound call

The license check sends the number of agents, the number of events and the server version. Never file names, commands, prompts or code.

R-Secure doesn't sit between your AI tools and their model providers. It stops the critical reads before the AI tool sees them, and records everything else. How data is handled →

R-Secure today

Four things it does now. Nothing it doesn't.

Capture, the kernel watcher, detection with blocking, and an audit log with a dashboard. Everything still in development is listed separately below.

01

Capture

Native hooks

Hooks into Claude Code, Cursor and Google Antigravity, and records every file read, edit, search, shell command and MCP tool call with its session, project and prompt. Prompt logging is optional.

LOGGEDCursor edited the file src/billing/invoice.tssession s412 · billing-web · hook
02

Kernel watcher

Linux · bpftrace

Hooks only see what the AI tool reports. The bpftrace watcher sees what it doesn't: scripts and subprocesses the agent starts, and files touched outside the hook path. These events are marked indirect and lead the daily report.

INDIRECTA script started by Claude Code read ~/.aws/credentialspython scripts/check_dag.py · kernel watcher · marco-linux
03

Detect & block

61 built-in rules

Every event is checked for secrets, cloud credentials, private keys, destructive commands, privilege escalation and more. For critical rules, the hooks stop the read or the command before it happens. Everything else is recorded and flagged.

BLOCKEDClaude Code tried to read ~/.ssh/id_ed25519KEY-001 · critical · stopped by the hook before the read
04

Audit log + dashboard

On your server

Events are kept in a SQLite log on your R-Secure server. The dashboard reads in plain English, and a daily report in HTML and Markdown gives you evidence for audits.

RECORDEDClaude Code read the file ~/payments-api/.envSEC-001 · fingerprint 3f9a1c2e · value not stored
Future development

We build and deliver 2× faster than our competitors.

  • Coming soon

    Warden

    Strip secrets and injected instructions before they reach the model.

  • Coming soon

    Sieve

    Check the packages an agent installs for malware and made-up names.

  • Coming soon

    Rules in the dashboard

    Edit rules in the console. Today they live in a config file on your server.

  • Coming soon

    More coding environments

    Setup for more AI coding tools and editors, coming to all platforms.

Who it's for

For teams with no MDM and no VPN.

Your developers bring their own Claude or Cursor subscriptions to company work. There's no device management on their laptops and no VPN in front of their traffic, so your security team can't see what those tools read or run.

R-Secure needs neither. Each developer runs one small agent that reports to a server you host.

  • 01

    Personal AI subscriptions

    The hooks run on the developer's machine, so it doesn't matter whose account the AI tool is signed in to.

  • 02

    No MDM required

    No device management or endpoint suite. A single-file agent sets up the hooks itself.

  • 03

    No VPN or proxy

    Traffic isn't rerouted. Developers keep working exactly as before, and if your server is down their AI tools still work.

Deploy

Self-hosted, on your machines.

Three steps, all on your infrastructure. There's no RAKSYON cloud to sign up for.

  1. Install the serverRun the R-Secure server on any machine in your network with your license key. The dashboard is ready immediately.
  2. Add developer machinesCreate a token for each machine. The developer runs the single-file agent, and it sets up the AI tool hooks automatically.
  3. Watch activity arriveEvents appear in the dashboard within seconds. Each seat covers one developer.
Hosting
Self-hosted, on your machines
Agents
Linux (Ubuntu), macOS, Windows
Kernel watcher
Linux only
If it's down
AI tools keep working normally
Example · r-secure feed · claude-code · payments-apilive
14:02:11readsrc/billing/invoice.tslogged 14:02:13read~/payments-api/.envcriticalSEC-001 secret value scanned in memory, fingerprint 3f9a1c2e kept 14:02:19execnpm test -- billinglogged 14:02:26read~/.ssh/id_ed25519blockedKEY-001 stopped by the hook before the read 14:02:31execpython scripts/sync.pylogged 14:02:33read~/.aws/credentialsindirectCLD-001 read by the script, seen by the kernel watcher 14:02:40writesrc/billing/invoice.ts +18 −4logged 14:02:45watching
How we compare

Lighter to run. Faster to help.

R-Secure installs in under 5 MB, uses a third of the memory, and a person answers when you need help.

Install size
<5MB

Against 200 MB on Windows and 60 MB on Ubuntu.

Windows

R-Secure<5 MB
Competitor200 MB

Ubuntu

R-Secure<5 MB
Competitor60 MB
Memory
3×

Less memory used

R-Secure uses a third of the memory of competing agents.

Support
10×

Faster human response

Your questions are answered by people on the RAKSYON team.

Pricing

Priced per developer. Talk to us for numbers.

Trial

For evaluating R-Secure with your own team.

Contact us
Start a trial
  • Try it with your team before you buy
  • Self-hosted, on your machines

Starter

Small teams

For a small team adopting AI coding tools.

Contact us
Talk to us
  • Organization dashboard
  • 61 built-in detection rules

Enterprise

For larger organizations.

Contact us
Talk to us
  • All features
  • More seats
  • Priority support
Every plan is self-hosted. One seat covers one developer.Prices, seat limits and plan features are being finalized.
Book a demo

Keep secrets inside your network.

A 30-minute walkthrough with the RAKSYON team, on your own infrastructure, with the AI tools your developers already use.

We reply from a real inbox.