Live in production · v1.4 · 76 tools

Give Claude
a brain,
nervous system
and hands.

The complete Claude Code setup running my live trading bot, six client projects, and autonomous agents. Deploy your own MCP server in ten minutes — your keys, your data, your server.

Browse 76 tools
One-time paymentLifetime updates7-day refund
AlpacaSupabaseGitHubFinnhubKrakenStripeMemoryVercelMCP76 TOOLSClaude Codeyour-server.vercel.app
Alpaca live tradingSupabase adminGitHub automationPersistent memoryFinnhub market intelReddit researchStripe opsSecurity auditsCRM + outreach12-turn orchestrationWordPress publishingKnowledge graphAlpaca live tradingSupabase adminGitHub automationPersistent memoryFinnhub market intelReddit researchStripe opsSecurity auditsCRM + outreach12-turn orchestrationWordPress publishingKnowledge graph
76tools
MCP server
300+
Skills library
8tier
Rule hierarchy
10min
To deploy
01 / Tools

Seventy-six tools.
One brain.

Every tool is registered across all required files — no silent 400s, no parameter mismatches. Search the full library below.

⌘K
B
get_context
Hydrate session with relevant memory
SUPERMEMORY
B
log_activity
Persist work outcome to memory bank
core
B
remember
Save fact/preference/decision
SUPERMEMORY
B
recall
Look up by file or function name
core
B
recall_context
Pull broader topical context
core
B
memory_recall
Direct memory key lookup
core
B
memory_search
Semantic search across all memories
core
B
memory_bank_get
Read structured memory bank entry
core
B
memory_bank_put
Write structured memory bank entry
core
B
memory_bank_search
Search structured memory bank
core
B
analyze_and_store
Auto-extract insights and persist
core
B
get_insights
Surface learned patterns
core
T
alpaca_get_account
Cash, buying power, equity
ALPACA
T
alpaca_get_positions
Live open positions
ALPACA
T
alpaca_get_orders
Pending and filled orders
ALPACA
T
alpaca_place_order
Market or limit, paper or live
ALPACA
T
alpaca_close_position
Exit at market
ALPACA
T
alpaca_get_clock
Market open status + next session
ALPACA
T
kraken_get_balance
Spot account balances
KRAKEN
T
kraken_place_order
Place spot crypto order
KRAKEN
T
kraken_cancel_order
Cancel pending order
KRAKEN
T
kraken_trade_history
Filled trades log
KRAKEN
T
trading_analysis
Multi-source signal synthesis
core
T
trade_engine_run
Run full strategy turn
core
T
crypto_signals
Momentum + risk signals
core
T
crypto_blitz
Short-window scanner
core
T
crypto_altblitz
Altcoin momentum scanner
core
T
crypto_strategy
Strategy backtest runner
core
T
kraken_signals
Kraken-derived signals
core
M
market_news
Aggregate top market headlines
FINNHUB
M
company_news
Per-ticker news feed
FINNHUB
M
stock_sentiment
Crowd + news sentiment score
FINNHUB
M
stock_quote
Live price + change
FINNHUB
M
market_intel
Quote + sentiment + news bundle
FINNHUB
M
technical_indicator
RSI, MACD, SMA on demand
FINNHUB
M
earnings_calendar
Upcoming earnings windows
FINNHUB
M
stock_signals
Composite buy/sell signal
core
G
github_list_repos
Owned + member repositories
core
G
github_read_file
Read any file at any ref
core
G
github_write_file
Commit a file change
core
G
github_create_issue
Open issue with labels
core
G
github_create_branch
Branch from main
core
G
github_create_pull_request
Open PR with body
core
G
github_create_issue_comment
Comment on issue or PR
core
S
supabase_query
Service-role SELECT bypassing RLS
core
S
supabase_insert
Service-role insert
core
S
supabase_list_tables
Schema inspection
core
S
security_audit
Full-stack vulnerability sweep
core
S
security_audit_github
Repo secrets, branch rules
core
S
security_audit_supabase
RLS, policies, exposed keys
core
S
security_audit_endpoints
Open routes, auth misses
core
S
security_audit_env
Env var hygiene check
core
C
find_leads
Source new lead candidates
core
C
save_lead
Persist lead to CRM
core
C
list_leads
Filterable lead list
core
C
update_lead
Update status, score, notes
core
C
crm_upsert_lead
Idempotent upsert
core
C
draft_outreach
Compose personalized email
core
C
send_outreach
Send via Resend
RESEND
R
reddit_search_summarize_store
Reddit DD with auto-store
core
R
google_ads_tracking_audit
GA4 + GAds tag check
core
R
wordpress_search
Site-wide WP content search
WORDPRESS
R
wordpress_publish
Publish/update WP post
WORDPRESS
R
stripe_get_customer_by_email
Lookup Stripe customer
STRIPE
R
stripe_list_invoices
Recent invoices for customer
STRIPE
G
knowledge_graph_add_edge
Connect two entities
core
G
knowledge_graph_neighbors
Traverse relationships
core
U
orchestrate
12-turn agentic loop, all tools
ANTHROPIC
U
calculate
Safe expression evaluator
core
U
generate_uuid
v4 UUID
core
U
get_current_time
ISO + tz aware
core
U
roll_dice
Notation parser
core
U
echo
Health check
core
U
test_notification
Push via ntfy
NTFY
U
list_decisions
Recent recorded decisions
core
U
orchestrate_status
Inspect running orchestration
core
76 of 76 toolsupdated daily · pull anytime
02 / The OS Layer

Rules built
from real failures.

Stripe double-charges. Auth bypasses. React stale state. Live trading disasters. Every line in the CLAUDE.md exists because something went wrong without it.

1# Ryan's AI Operating System
2 
3## TIER 1: ALWAYS READ FIRST
4 
5### MCP IS THE AUTHORITY
6- MANDATORY FIRST ACTION every session:
7 call get_context with topic
8- Do NOT respond until called + read
9- Call recall before editing
10- If MCP says X, follow X. If training says Y, ignore Y.
11 
12### READ BEFORE ACT — NO EXCEPTIONS
13- ALWAYS read file before editing
14- NEVER generate for unread functions
15 
16### TYPE-CHECK AFTER EVERY TS CHANGE
17- Run `tsc --noEmit` — no exceptions
18 
19### AUTONOMY — NEVER ASK PERMISSION
20- NEVER tell user to do manually
21- "Check X" = diagnose + fix + report
22 
23# Built from real failures.
24# Every rule = something that broke.
F.01
Auto-memory hooks

Pre/post-task hooks save user feedback, project context, and decisions automatically. Every session restores summary on start.

  • Self-learning pattern detection
  • Fact-Forcing Gate on destructive commands
  • Post-edit quality gate triggers code review
F.02
Skill auto-triggers

30+ patterns invoke the right skill automatically — no prompting, no remembering. "Review this" calls the reviewer; "deploy" runs the verification stack.

  • everything-claude-code skill pack
  • SPARC: architect, coder, TDD, security
  • Vercel deploy → verification chain
F.03
Definition of Done

Hard checklist runs before any task is marked finished — type-check, build, secrets, log_activity. No more "looks good to me" regressions.

  • tsc --noEmit zero tolerance
  • npm test + npm run build green
  • log_activity called on every result
F.04
Multi-agent verification

For payments, auth, and booking logic — a second agent independently verifies the change. No "looks right" on security paths.

  • Pre-check: read all connected files
  • Post-check: verify adjacent code intact
  • 12-turn orchestration loop
03 / Deploy

Your server.
Your keys.

Push the GitHub repo, click Deploy, paste your env vars. Ten minutes from purchase to a private MCP server humming on Vercel.

Deploy to Vercel
~3 min · 1 click
Cloning repo…
Installing deps…
Building Next.js…
Deploying functions…
✓ Live in 47s
ryanbach91 / claude-os
private · main
76 tools300+ skillsv1.4 latest● live
CLAUDE.mdglobal AI OS — 6 tier rules
mcp-server/76+ tools, ready to deploy
skills/300+ skill library
hooks/auto-memory + quality gate
scripts/install.jsone-command setup
templates/project CLAUDE.md template
04 / Pricing

One price. Everything.

No subscription. No upsells. No “starter tier.” The full setup, lifetime updates, 7-day refund.

One-time
$19.99
Less than dinner. Pays for itself in one prompt.
secure checkout · stripe · 7-day refund
What you get
  • Global CLAUDE.md — the AI OS
  • MCP server — full source, 76 tools
  • Skill library — 300+ skills
  • Hooks & auto-memory system
  • Project CLAUDE.md template
  • One-command install script
  • Vercel deploy guide + env template
  • Lifetime GitHub repo access
  • All future updates
  • 7-day no-questions refund
05 / FAQ

Things people ask.

Yes. If you can copy-paste a command and click "Deploy" on a website, you can run this. The install script copies everything to ~/.claude/ in one command. The MCP server deploys to Vercel with a button click. You connect your own API keys (most have free tiers) and Claude Code instantly becomes 10× more capable. No coding required to use it — just to extend it later if you want.
It's a persistent brain for Claude. Without it, Claude forgets everything between conversations and can't touch your data. With it, Claude remembers your projects, reads your databases, manages your GitHub, places live trades, audits your security — all autonomously. It's the difference between a chatbot and an actual assistant.
Claude Code out of the box is generic. It uses ~80% of its context on basic instructions you have to repeat every session. This setup hands Claude a structured 6-tier rule system, automatic memory, and 76 tools — so your prompts go from "please remember to type-check" to "ship the feature." One pro user reported their average session got 3× longer in useful output.
Tools support both. Default is paper (fake money, real market data). Flip ALPACA_PAPER=false to go live. The CLAUDE.md has hard rules — −8% stop loss, +15-20% take profit, always check market clock first — built from real losses so you don't repeat them.
No. You deploy your own instance to Vercel with your own keys. Your data stays on your server. My production server stays mine. You get the full source code — fork it, extend it, white-label it for clients.
7-day refund, no questions, just email ryan.bach91@gmail.com. The whole package is $19.99 one time — less than dinner.
Yes. Lifetime access to the GitHub repo. New tools, new skills, new patterns — pull whenever you want.
Still on the fence?

Email me. I'll answer in a day. ryan.bach91@gmail.com

Claude OS — full setup$19.99