Cavaridge AgentOps Kit
Autonomous AI workers your MSP can actually trust.
An implementation kit for MSPs and healthcare IT operators: AI workers with persistent memory, approval gates, auditability, and safe, scoped tool access — built on Claude Code, Telegram, and plain-markdown memory. Not a chatbot.
One-time purchase. Instant download. Done-for-you setup available.
You run the operation. The busywork runs you.
MSP and IT teams are buried in repeatable work — and the usual AI advice is either toy automation or a security incident waiting to happen.
Repeatable ops eat senior time
Ticket triage, status reports, runbook steps, client updates. Necessary, low-leverage, and constant — and it's your most expensive people doing it.
Knowledge lives in people's heads
Runbooks, client quirks, and the decision rules that took years to learn are uncaptured and unauditable — and they walk out the door when someone's on PTO.
Un-auditable automation is a liability
In regulated and healthcare environments, “the script did it” isn't an answer. You need approval gates and an audit trail, not a black box.
Broad AI access is how incidents start
An agent holding your credentials with a browser open is a breach surface. Scoped tool access and explicit gates aren't optional at operator scale.
What's included
A practical, implementation-focused system for operators — memory, approval gates, and a security model you can audit. Not theory, not hype.
AgentOps Toolkit
$99
One-time
- ✓Agent memory architecture (PARA, git-versioned audit trail)
- ✓Workflow & automation templates
- ✓Telegram command channel with owner auth
- ✓Approval gates & trust model
- ✓Scoped tool-access patterns
- ✓Prompt library
- ✓PDF + Obsidian Markdown
Implementation Bundle
$299
One-time
- ✓Everything in the Toolkit
- ✓Repo scaffold + PARA init script
- ✓Cron jobs: nightly consolidation + heartbeat
- ✓Telegram bot template (Node)
- ✓Security model & scoped-credential setup
- ✓30-day rollout checklist
- ✓Private operator community
Need it built for you? The Done-For-You Setup delivers a private operator, configured and handed off. See pricing.
How the system is structured
The security model is the foundation, not an afterthought. Trusted command channels are explicitly separated from untrusted data, and every action is scoped, gated, and logged.
Command Channel (Authenticated)
- Telegram owner/admin messages
- Local operator terminal
Agent Core
- Claude Code / Codex session
- PARA memory (markdown files)
- Daily notes writer
- Project tracker
- Cron scheduler
Delegated Work
- Long-running coding tasks
- Nightly memory consolidation
- Heartbeat / health checks
- Web research sub-agents
Information Channels (Untrusted)
- Web pages
- Uploaded documents
- Public comments / social
Financial, DNS, production deploy, and destructive actions require explicit owner approval.
Free excerpt
Isolation & scoped tool access
## Why least privilege matters
Every AI worker should run with minimal privilege: a separate OS user account, a separate scoped API key per integration, and no access to your primary browser session, password manager, or production credentials.
The goal isn't paranoia — it's containment. If a worker makes a mistake or gets manipulated by injected instructions in a web page, ticket, or email, the blast radius stays inside a sandboxed scope. That's the difference between an automation and an incident.
## Setup checklist
- →Create a dedicated macOS or Linux user account (e.g. `agent`)
- →Generate a new SSH key pair for the agent — do not reuse existing keys
- →Create a separate GitHub account or use a machine user with scoped repo access
- →Create Telegram bot via @BotFather — store token in agent env only
- →Use a dedicated email address for all agent-created accounts
- →Never grant the agent access to your primary Stripe, AWS, or GCP account
- →Use read-only API keys wherever write access is not required
The Implementation Bundle includes environment setup scripts, .env templates, and a credential rotation schedule.
Pricing
From a do-it-yourself toolkit to a private operator we build for you.
AgentOps Toolkit
$99
One-time
The reference kit: give an AI worker durable memory and safe, auditable workflows.
- ✓ Agent memory architecture (PARA, git-versioned audit trail)
- ✓ Workflow & automation templates
- ✓ Telegram command channel with owner auth
- ✓ Approval gates & trust model
- ✓ Scoped tool-access patterns
- ✓ Prompt library
- ✓ PDF + Obsidian Markdown
Implementation Bundle
$299
One-time
Everything to stand it up: scaffold, scripts, and the full security model.
- ✓ Everything in the Toolkit
- ✓ Repo scaffold + PARA init script
- ✓ Cron jobs: nightly consolidation + heartbeat
- ✓ Telegram bot template (Node)
- ✓ Security model & scoped-credential setup
- ✓ 30-day rollout checklist
- ✓ Private operator community
Done-For-You Setup
$1,500–$5,000
$1,500 refundable deposit to start
We stand up a private, autonomous operator for your shop — memory, approval gates, audit logs, and a clean handoff.
- ✓ Private autonomous operator, built for you
- ✓ Scoped tool access + approval gates
- ✓ Audit-ready memory & action logs
- ✓ MSP/client-ready handoff & docs
- ✓ $1,500 refundable deposit, credited to your project
Cavaridge AgentOps
SaaS
Control plane — coming soon
Hosted agent memory and a control plane for running fleets of operators with audit and approval built in.
- ✓ Hosted agent memory & daily synthesis
- ✓ Auditable actions & approval workflows
- ✓ Scoped tool access (RBAC)
- ✓ Per-tenant isolation
- ✓ Team & client management
Payments processed by Stripe. Secure checkout. Instant access for download tiers.