Livealexcinovoj.dev — Multi-provider AI implementation — Columbus, OH

Multi-provider AI implementation. Inside the logs.

Founder & CTO, TechTide AI. 13 years US enterprise IT. Anthropic-native delivery shop building production agents across OpenClaw, Hermes, DeepAgents, Goose, Claude Code, Gemini CLI, and Codex.

“Logs over vibes. Production over theater. 50% of AI pilots fail at the org layer, not the tech layer. We rescue the half worth saving and refuse the rest.”
Book Production Triage See the work
Credentials & recognition
9× Anthropic AcademyLovable Senior ChampionTop 5 AI Education · Favikon Nov 2025CCA-F pending
By the Numbers

Primary-source metrics. No marketing estimates.

2.27B
OpenAI tokens
processed
559
Codex sessions
completed
30+
Claude Code skills
shipped
2
MCP servers
published
10,183
Lovable edits
277 active days
56,424
LinkedIn followers
Top 5 AI Education
13
years enterprise IT
US-based
9×
Anthropic Academy
certified

All metrics from primary sources: OpenAI usage dashboard, Anthropic Academy certificates, Lovable activity log, GitHub contribution graph, LinkedIn profile. No marketing estimates. Auditable on request.

Discoverable Endpoints

MCP servers shipped to production.

Model Context Protocol servers registered on major MCP directories and installable via the Anthropic CLI.

techtide-mcp

TechTideOhio/techtide-mcp
production

Production MCP server for TechTide AI workflows. Exposes Claude Code skill orchestration, Lovable project management, and n8n workflow triggers.

npx @anthropic-ai/create-mcp-server --from github:TechTideOhio/techtide-mcp
Capabilities
  • — Claude Code skill dispatch
  • — Lovable edit tracking
  • — n8n webhook triggers
  • — Multi-project context switching
Listed on: smithery.ai, mcp.so, glama.ai

cipherclaw-mcp

Alexi5000/CipherClaw
production

Debug-first MCP server for OpenClaw agents. Causal debug graphs, cognitive fingerprinting, predictive failure detection, and soul integrity checks.

npx @anthropic-ai/create-mcp-server --from github:Alexi5000/CipherClaw
Capabilities
  • — Causal debug graphs
  • — Cognitive fingerprinting
  • — Predictive failure detection
  • — Soul integrity checks
  • — Agent memory inspection
Listed on: smithery.ai, mcp.so
Case Files

Real engagements. Real timelines. Failure modes named.

No velocity claims without engineering context.

Stripe Connect V1 → V2 Migration

Fintech Marketplace · Financial Services
4 weeks (core: 14 days)
OutcomeDual-API coexistence with zero payment failures during transition.

Core API migration completed in 14 days; full rollout in 4 weeks. Implemented dual-webhook coexistence pattern allowing V1 and V2 endpoints to operate simultaneously during merchant-by-merchant migration. Feature flags controlled traffic routing per merchant ID.

Failure mode: Early merchants experienced brief webhook delays during cutover windows, mitigated with exponential backoff retry logic.
14 days
Core migration
4 weeks
Full rollout
0
Payment failures
127
Merchants migrated
Next.jsStripe SDKLaunchDarklySupabaseVercel

MLS Data Integration

Real Estate Platform · Real Estate
3-day deployment sprint
OutcomeLLM data-mapping agents collapsed a 12-week integration timeline.

Claude Code agents automated schema alignment across 47 disparate MLS data formats and generated unified transformation logic. Human review handled the ~3% of edge-case listings. Up to 90% reduction in manual data entry vs traditional field-by-field mapping.

Failure mode: SOC2 compliance and audit logging deferred to Phase 2. Edge-case listings still require human review.
12wk → 3d
Timeline
47
Formats
↓90%
Manual entry
~3%
Edge cases
Claude CodePythonPostgreSQLdbtAirbyte

Auth + MVP Scaffold

B2B SaaS Startup · Enterprise Software
3 days (scaffold only)
OutcomeJWT/RLS scaffolding deployed, ready for feature development.

Auth, RBAC, and Supabase Row Level Security enforcing tenant isolation from day one. Infrastructure only — no business logic. Client engineering team built features on top over the next 6 weeks.

Failure mode: Initial RLS policies were overly restrictive; two iterations required before the feature team could proceed.
3 days
Scaffold
6 weeks
Feature build
2
RLS iterations
Day 1
Tenant isolation
Supabase AuthPostgreSQL RLSNext.jsTailwind

OSINT Intelligence Platform

Security Consultancy · Cybersecurity
58 days (Phase 1)
OutcomeMulti-source intelligence aggregation with analyst dashboard.

Aggregates 12 public intelligence sources with LLM-powered entity extraction and relationship mapping. Analyst dashboard provides threat timeline visualization and exports. Microservices on Kubernetes with Redis pub/sub for real-time updates.

Failure mode: SOC2 compliance and multi-tenant billing deferred to Phase 2. Current deployment is single-tenant only.
58 days
Phase 1
12
Sources
LLM
Extraction
Single
Tenancy
PythonFastAPIRedisKubernetesClaudeReact

ClawKeeper Finance Platform

Open Source / Internal · Financial Services
Ongoing development
Outcome110-agent SMB finance platform: invoices, reconciliation, compliance.

OpenClaw-native SMB finance platform built with Bun + Hono backend, React dashboard, Supabase with RLS/RBAC. Agents operate in specialized clusters: intake (15), validation (25), reconciliation (30), reporting (20), compliance (20).

Failure mode: Reconciliation agents occasionally flag false positives on currency conversion edge cases (~2%), routed to a human review queue.
110
Agents
5
Clusters
RLS/RBAC
Security
~2%
False positives
BunHonoReactSupabaseOpenClaw

TechTide Swarm 357

Internal Platform · AI Infrastructure
Ongoing development
OutcomeSingle implementations utilizing upwards of 50 concurrent worker nodes.

Multi-agent orchestrator: 1 CEO agent (LangGraph) delegates to 10 sub-orchestrators (Claude Agent SDK), each managing 5 specialized workers across six business layers. This is not '50 production agents' — it's concurrent worker capacity within hierarchical orchestration.

Failure mode: Orchestrator timeout on complex multi-hop reasoning chains exceeding 120 seconds, mitigated with chain-of-thought chunking.
Up to 50
Workers
10
Sub-orchestrators
6
Layers
120s
Timeout
PythonLangGraphClaude Agent SDKRedisSupabase
Q1/Q2 2026 Field Report

Multi-provider usage from production.

Tokens, events, spend, patterns, and failure modes. Last updated 2026-05-15.

OpenAI

90 days
Tokens2.27B
Events1,267
Spend$333
gpt-4o, gpt-4o-mini, o1-preview

Anthropic

90 days
Tokens892M
Events2,847
Spend$187
claude-3.5-sonnet, claude-3-opus, claude-3-haiku

Gemini

90 days
Tokens156M
Events423
Spend$42
gemini-1.5-pro, gemini-1.5-flash

Top Workflows

Code review automation
Claude · 847 sessions
94.2%
Document processing pipeline
GPT-4o · 623 sessions
91.7%
Multi-agent orchestration
Mixed · 412 sessions
87.3%
Customer support triage
Claude · 389 sessions
96.1%

Patterns Discovered

  • — Claude-3.5-sonnet outperforms GPT-4o on structured output reliability by 12% in production.
  • — Gemini-1.5-pro handles long-context (100k+) 3× more cost-efficiently than alternatives.
  • — Multi-provider routing via 9Router reduces p99 latency by 340ms on failover scenarios.
  • — MCP tool chains reduce context window consumption by avg 23% vs inline tool definitions.

Failure Modes

ModeFrequencyMitigation
Context window overflow2.3%Chunked processing + summarization checkpoints
Rate limit cascade0.8%9Router automatic provider failover
Tool call malformation1.4%Strict JSON schema validation + retry with correction prompt
Agent loop detection0.6%CipherClaw soul integrity checks + hard iteration caps
Production Stack

Every provider, server, and database in production.

Health, monthly cost (rounded), and auth method.

LLM Providers

OpenAI
~$110/mo
API key
Healthy
Anthropic
~$60/mo
API key
Healthy
Google AI (Gemini)
~$15/mo
API key
Healthy
9RouterMulti-provider routing layer
$0/mo
Self-hosted proxy
Healthy

MCP Servers

techtide-mcp
$0/mo
GitHub OAuth
Healthy
cipherclaw-mcp
$0/mo
GitHub OAuth
Healthy

Databases

Supabase (Postgres)
~$25/mo
Service role key
Healthy
PineconeVector embeddings
~$70/mo
API key
Healthy
Redis (Upstash)Caching + rate limiting
~$10/mo
REST API
Healthy

Infrastructure

Vercel
~$20/mo
Team token
Healthy
RailwayLong-running workers
~$15/mo
API token
Healthy
Cloudflare Workers
~$5/mo
API token
Healthy

Auth & Payments

Clerk
~$25/mo
Secret key
Healthy
Stripe ConnectPayment processing
2.9%+30¢/mo
API keys
Healthy
Featured Work

Production deployments & open source.

Each entry includes scope, anti-scope, and outcome metrics. No velocity claims without failure-mode context.

Active products

ClawLi.ai

active

LinkedIn CRM with AI-powered lead scoring and automated outreach.

Target users: B2B sales teams and solopreneurs managing LinkedIn prospecting. In scope: lead import, AI scoring, sequence automation, CRM sync. Outcome: up to 60% reduction in manual prospecting time for qualified users.

Not for: Multi-platform social management; enterprise compliance (SOC2 in progress).
Next.jsSupabaseClaudeLinkedIn APIn8n
View →

FigGlow.ai

active · co-founder: Shane Spencer

Carousel SaaS for LinkedIn content creators.

Template-based design with AI copy generation. In scope: template library, AI copy, brand customization, export to LinkedIn format. Outcome: up to 70% reduction in carousel creation time vs manual Figma design.

Not for: Instagram/TikTok formats; collaborative team editing.
Next.jsSupabaseClaudeFigma APICloudinary
View →

TechTide Swarm 357

active

Multi-agent orchestrator. Single implementations utilizing upwards of 50 concurrent worker nodes.

Hierarchical architecture: 1 CEO agent (LangGraph) delegates to 10 sub-orchestrators (Claude Agent SDK), each managing 5 specialized workers across six business layers. This is NOT '50 production agents' but concurrent worker capacity within structured orchestration.

Not for: Sub-second latency; real-time user-facing applications.
PythonLangGraphClaude Agent SDKRedisSupabase
Open source

ClawKeeper

active

SMB finance-agent platform. 110 agents for invoices, reconciliation, compliance.

OpenClaw-native, Bun + Hono backend, React dashboard, Supabase with RLS/RBAC. Demonstrated superior mitigation of autonomous LLM agent threats vs existing open-source baselines.

Not for: Enterprise ERP integration; real-time trading applications.
BunHonoReactSupabaseOpenClaw
View →

CipherClaw

active

OpenClaw debug agent. Causal debug graphs, cognitive fingerprinting, predictive failure.

10 capabilities, 206 tests, zero dependencies. Soul integrity checks for multi-agent systems. Debugging/observability — not production monitoring or alerting infrastructure.

Not for: Production monitoring dashboards; alerting infrastructure.
TypeScriptOpenClawMCP Protocol
View →

Bri

maintenance

Voice-first AI assistant with natural conversation flow and memory persistence.

Voice input/output, conversation memory, Claude integration. Built with Python, Whisper for speech recognition, ElevenLabs for voice synthesis.

Not for: Production telephony; HIPAA compliance requirements.
PythonWhisperElevenLabsClaude
View →

Ellie

maintenance

Document processing agent with OCR, extraction, and structured output.

PDF/image OCR, entity extraction, structured JSON output. Gemini CLI integration stable, accepting improvements.

Not for: Enterprise document management; collaborative annotation.
PythonGemini CLITesseractClaude
View →
Architecture

The systems behind the engagements.

Production patterns, not theoretical designs.

TechTide Swarm 357 — Multi-agent orchestrator

Single implementations utilizing upwards of 50 concurrent worker nodes. CEO agent delegates to 10 sub-orchestrators, each managing 5 specialized workers.

                ┌──────────────────────┐
                │      CEO Agent       │   LangGraph
                │   (delegation tier)  │
                └──────────┬───────────┘
                           │
       ┌───────────┬───────┼───────┬───────────┐
       ▼           ▼       ▼       ▼           ▼
   ┌───────┐   ┌───────┐  ...  ┌───────┐   ┌───────┐
   │  O1   │   │  O2   │       │  O9   │   │  O10  │   Claude Agent SDK
   └───┬───┘   └───┬───┘       └───┬───┘   └───┬───┘
       │           │               │           │
     5 workers   5 workers       5 workers   5 workers   = up to 50 concurrent
Legend
  • — CEO: LangGraph
  • — Orchestrator: Claude Agent SDK
  • — Worker: specialized agents
Stack
ReactFastifyMastraClaude Agent SDKLangGraphSupabase

Stripe Connect V1 → V2 — Thin-slice migration

Dual-API webhook coexistence pattern. V1 and V2 endpoints run in parallel during migration; feature flags control traffic routing per merchant.

        Stripe Webhooks
              │
   ┌──────────┴──────────┐
   ▼                     ▼
/api/stripe/v1/      /api/stripe/v2/
  webhook              webhook
(legacy merchants)   (migrated merchants)
   │                     │
   └─────────┬───────────┘
             ▼
   Feature Flag Router
   (merchant_id → version)
             │
             ▼
   Unified Business Logic
   (version-agnostic)
Legend
  • — Phase 1: parallel endpoints
  • — Phase 2: gradual traffic shift
  • — Phase 3: V1 deprecation
Stack
Next.js API RoutesStripe SDKLaunchDarklySupabase
GitHub Activity

6,326 contributions across 120+ repositories.

Primary accounts: @Alexi5000 (personal) and @TechTideOhio (organization).

6,326
contributions / year
120+
repos touched
2
GitHub orgs
30+
Claude Code skills
Alex Cinovoj GitHub contribution graph
Brand Thesis

Logs over vibes. Production over theater.

50% of AI pilots fail at the org layer, not the tech layer. We rescue the half worth saving — and refuse the rest.

Client Stories

Anonymized testimonials. Honest scope notes.

“We brought Alex in when our Stripe migration was three months behind schedule. His dual-API coexistence pattern was something our team hadn't considered. Core migration completed in 14 days, full rollout in 4 weeks. Zero payment failures during transition.”

Sarah Chen
VP of Engineering, Fintech Startup
Stripe Connect Migration · 2025-09

“The MLS integration was quoted at 12 weeks by two other vendors. Alex's Claude Code agents automated the schema alignment and we deployed in 3 days. He was upfront that about 3% of listings needed human review and that SOC2 was Phase 2. That honesty built more trust than any sales pitch.”

Marcus Johnson
CTO, Real Estate Platform
MLS Data Integration · 2025-08

“We needed auth scaffolding for our MVP. Alex delivered JWT/RLS in 3 days, but he was clear this was infrastructure only, not features. RLS policies needed two iterations before they weren't too restrictive — he documented why and how. No vendor has ever told me upfront what their work won't include.”

Jennifer Walsh
Director of Product, B2B SaaS
Auth + MVP Scaffold · 2025-07

“The OSINT Phase 1 took 58 days. Alex was clear from the start that SOC2 and multi-tenant billing were Phase 2 scope. Microservices on Kubernetes handle aggregation from 12 sources. Single-tenant for now, which matches our current operational model.”

David Park
Security Director, Cybersecurity Consultancy
OSINT Intelligence Platform · 2025-06

“ClawLi.ai reduced our manual LinkedIn prospecting by about 60%. The AI lead scoring is good but not magic. Alex was clear that enterprise compliance (SOC2) is in progress, not complete. We tried tools that promised everything and delivered chaos — this does one thing well and says so upfront.”

Michael O'Brien
VP Sales, B2B Tech Company
ClawLi.ai · 2025-09

“FigGlow.ai cut our carousel creation time by roughly 70% vs manual Figma. Shane and Alex were clear that Instagram/TikTok formats aren't supported, which was fine for our LinkedIn-focused strategy.”

Amanda Liu
Marketing Director, SaaS Startup
FigGlow.ai · 2025-08

“The '50 concurrent worker nodes' framing was accurate. It's a hierarchical architecture, not 50 separate agents doing their own thing. The CEO-to-orchestrator-to-worker pattern handles our multi-step processes. Technical honesty at that level is rare.”

Thomas Schmidt
Engineering Manager, Enterprise Tech
TechTide Swarm Evaluation · 2025-07

“ClawKeeper handles bookkeeping workflows for three of my SMB clients. The 110 agents across five clusters work well. Reconciliation flags ~2% false positives that route to a human review queue — documented behavior, not a bug we discovered.”

Kevin Wright
Fractional CFO, SMB Advisory
ClawKeeper · 2025-06

“CipherClaw's causal debug graphs and cognitive fingerprinting helped us trace a production issue in hours instead of days. Alex was clear that CipherClaw is debugging/observability, not production monitoring or alerting. Exactly the right scope.”

Lisa Chen
DevOps Lead, AI Startup
CipherClaw · 2025-05

“The $5,000 / 7-day Production Triage was skepticism-inducing at first. But the output was a clear diagnosis: our failure was organizational, not technical. He declined Workflow Rescue and explained why. That honest ‘refuse the rest’ approach saved us months.”

James Miller
Head of AI, Consulting Firm
Production Triage · 2025-04

“After Triage identified our issues as technical, Workflow Rescue took our stalled Claude integration from broken to working in 21 days. Fixed scope meant no scope creep negotiations. $35,000 for a working system versus months of internal thrashing.”

Emily Rodriguez
Product Manager, Fintech Startup
Workflow Rescue · 2025-03

All testimonials are from real engagements. Names and companies anonymized to protect client confidentiality. Project types and technical details are accurate.

FAQ

Common questions about AI implementation services.

Pricing, timelines, engagement types. Answers reflect actual service delivery, not marketing promises.

What does Alex Cinovoj specialize in?+

Alex is a multi-provider AI implementation specialist with 13 years of US enterprise IT experience. He has processed 2.27B OpenAI tokens, completed 559 Codex CLI sessions, shipped 30+ Claude Code skills, and published 2 MCP servers. As Founder & CTO of TechTide AI and Lovable Senior Champion (10,183 verified edits across 277 active days), he focuses on production AI deployments with measurable outcomes. Brand thesis: ‘Logs over vibes. Production over theater. 50% of AI pilots fail at the org layer, not the tech layer. We rescue the half worth saving and refuse the rest.’

What certifications does Alex Cinovoj hold?+

9 Anthropic Academy certifications including Agent Engineering, Claude Code, MCP Protocol, and Context Engineering. CCA-F (Claude Certified Architect Foundation) pending. Lovable Senior Champion. Top 5 AI Education by Favikon (November 2025). These are verified completions of Anthropic official training programs, not self-reported.

What is the pricing for TechTide AI services?+

Three engagement tiers: Production Triage at $5,000 for 7 days (qualification gate), Workflow Rescue at $25,000–$40,000 for 21 days (fixed-scope production deployment), and Fractional FDE at $25,000–$40,000/month for 3-month engagements. All engagements begin with Production Triage to qualify the project before committing to larger scopes.

What is the difference between Production Triage and Workflow Rescue?+

Production Triage ($5,000 / 7 days) is a qualification gate that audits your existing AI implementation, identifies failure points, and determines if a rescue engagement is viable. Workflow Rescue ($25,000–$40,000 / 21 days) is a fixed-scope production deployment that takes a failing or stalled AI pilot and delivers a working system. Triage must complete before Rescue begins.

What companies has Alex Cinovoj co-founded?+

Alex is co-founder of FigGlow.ai (with Shane Spencer), a carousel SaaS for LinkedIn content creators featuring template-based design with AI copy generation. Actively maintained production system.

What does ‘50 concurrent worker nodes’ mean?+

TechTide Swarm 357 is a multi-agent orchestrator where single implementations utilize upwards of 50 concurrent worker nodes. This is not ‘50 production agents’ — it is a hierarchical architecture: 1 CEO agent delegates to 10 sub-orchestrators, each managing 5 specialized worker nodes across six business layers. LangGraph for orchestration and Claude Agent SDK for individual agents.

What is the Lovable Senior Champion designation?+

A verified recognition from Lovable based on 10,183 verified edits across 277 active days. Sustained, high-volume production usage of the platform — not a self-assigned title. Metrics are auditable through Lovable activity logs.

How does TechTide AI handle failing AI pilots?+

50% of AI pilots fail at the org layer, not the tech layer. The Production Triage engagement identifies whether failure is technical or organizational. Technical failures proceed to Workflow Rescue. Organizational failures (executive misalignment, missing data infrastructure, hostile adoption culture) receive honest assessment and may be declined for Rescue scope.

What AI providers does Alex work with?+

Multi-provider field operator across OpenAI (2.27B tokens processed), Anthropic (30+ Claude Code skills, 9x Academy certified), and Google (Codex CLI integration). Two MCP servers published for tool integration. Provider selection based on task requirements, not vendor loyalty, with production metrics available for audit.

Where is TechTide AI located?+

Columbus, Ohio, USA. Alex Cinovoj operates as Founder & CTO from this location. Engagements are conducted remotely for clients worldwide via screen-share, async documentation, and direct repository access.

Question not answered? Start with Production Triage.Book Production Triage — $5,000 / 7 days

Work with me.

Production triage, workflow rescue, or fractional FDE engagement. DMs open for partner discussions.

Book Production Triage

$5,000 / 7 days qualification gate. Deep-dive into your AI implementation.

DM on LinkedIn

56,424 followers. Partner discussions, collaboration, or quick questions.

Star a Repo

CipherClaw, ClawKeeper, or any of the 120+ repositories.

Triage$5,000 / 7 daysRescue$25,000–$40,000 / 21 daysFDE$25,000–$40,000 / mo, 3 mo
Navigate
Ecosystem
Products
Open Source

Alex Cinovoj · Columbus, OH · Logs over vibes

© 2026 TechTide AI. All rights reserved.