convez.ai
ComandoAutomaçãoIntermediário
avelikiy

Great CTO

avelikiy/great_cto

Autopilots de IA que executam processos de negócio end-to-end com humanos nas decisões críticas

Ver no GitHub
BomPluginScore · de 0 a 100

Command-line que inicializa autopilots de IA para funções empresariais completas (medical-coding, legal, accounting, etc). Detecta automaticamente sua stack, seleciona o arquétipo correto e bootstraps a documentação do projeto em um único comando.

Instalação

$npx convez-ai avelikiy/great_cto

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
117
Forks
12
Acompanhando
1
Saúde
85%
Último commit
há 2 dias
Versão
v4.1.0
Issues abertas
10
Contribuidores
3

Casos de uso

  • Automatizar fluxos de negócio complexos (medical-coding, bookkeeping, legal-document)
  • Implementar sistemas com escalação inteligente para decisões de risco
  • Estruturar projetos de IA enterprise com compliance e conectores reais

Veredito

Ferramenta amadura para implementar autopilots de IA em processos empresariais reais, com foco em resultado (redução de custos/tempo comprovada). Exige expertise em integrações e compliance, mas elimina repetição em setup.

README

A tradução para o português deste README ainda está na fila.

AI autopilots for business — get the work done, not just the software.

npm npm downloads License Claude Code Plugin Codex Savings

npx great-cto init

Website · One real run → · Live demo · Discussions · Changelog

Русский · 简体中文 · 繁體中文 · 日本語 · 한국어 · Español · Português · Deutsch · Français


Services are the new software

The next wave isn't tools for specialists — it's autopilots that sell the outcome of a service. An autopilot runs a whole business function end to end (intake → process → decide → deliver) and escalates only the judgment calls to a qualified human. Every model improvement makes the service faster and cheaper.

GreatCTO ships those autopilots — each one a flow of agents + tools with a human on the risky steps, a built-in compliance reviewer, and live connectors that run each flow on real data.

The autopilots

AutopilotWhat it doesMarketWho's building it
🩺 Medical-codingClinical notes → clean, compliant claims; a certified coder signs the risky ones$50–80BAnterior · CodaMetrix · Fathom
🖥️ Managed-ITPatches, configs & access across the fleet — staged, reversible, human on big changes$100B+Serval · Edra · Electric AI
⚖️ Legal-documentDrafts & redlines contracts and NDAs; a licensed attorney signs anything that's advice$20–25BCrosby · Harvey · Robin AI
📒 Bookkeeping & closeBooks, reconciles & closes the month; a controller signs the close$50–80BRillet · Basis · Digits
🧾 Tax-prepPrepares returns & classifies positions; a credentialed preparer signs before filing$30–35BBlack Ore · April · Column Tax
🛒 Source-to-payOnboards suppliers, matches invoices, releases payments — screened for sanctions & fraud$200B+Tacto · Zip · AskLio
✅ Prior-authorizationAuth request + chart → approval or a clean determination; a medical director signs every denial$35–56BCohere Health · Anterior · Develop Health
🛡️ KYC/AMLOnboards, screens & monitors customers; a BSA Officer signs every SAR$61BDiligent AI · Alloy · Sardine
🔐 Managed-SOCTriages & investigates every alert 24/7; a SOC analyst signs any containment$4–6B7AI · Dropzone · Prophet Security
☂️ Claims & underwritingAdjudicates claims & prices risk; a licensed adjuster/underwriter signs the call$36–38BShift · Akur8 · Avallon
🏠 Mortgage-underwritingProcesses & underwrites to clear-to-close; a DE underwriter signs$40B+Tidalwave · Zest AI · Blend
🔑 Title & escrowTitle search, escrow & closing; a licensed officer signs the title & the wire$16.2BPropy · Qualia · Titl
🪪 Provider-credentialingPrimary-source verifies & enrolls a provider; the committee signs privileging$1.2B+Medallion · Verifiable · CertifyOS
💰 Debt-collectionCompliant outreach & payment plans; a manager signs escalation & settlements$13.5–16BCollectWise · InDebted · Tesorio
🚚 Freight-brokerageMatches loads to vetted carriers & books; a licensed broker signs binding rates$19–125BHappyRobot · FleetWorks · Vooma
🧪 Clinical-trial-opsScreens patients & runs monitoring; the PI / medical monitor signs eligibility$20–28BTriomics · QuantHealth · Tempus
🛃 Customs-clearanceClassifies, values & screens imports; a licensed customs broker signs the entry of record$4.6BDigicust · Flexport · Avalara
📊 SOX ITGC auditTests IT general controls & drafts workpapers; a CPA signs the opinion$15–25BMidship · Scytale · AuditBoard
💊 PharmacovigilanceProcesses adverse-event cases; a QPPV / drug-safety physician signs before reporting$1.65BArisGlobal · Indegene · Lifebit
🗽 ImmigrationAssembles petitions & checks eligibility; a licensed attorney of record signs before filing (UPL)$3.2BBoundless · Lawfully · Docketwise
🏠 AppraisalPulls comps & reconciles value; a state-certified appraiser signs every USPAP report (independence)$9.2BClear Capital · Reggora · Class Valuation
💵 PayrollComputes gross-to-net & tax; a payroll manager (CPP) signs before ACH funding + the 941 deposit$5.8BGusto · Rippling · Deel
🦺 Workers-compDetermines compensability & benefits; a licensed adjuster signs denials & terminations (bad-faith)$6.5BCLARA Analytics · Gradient AI · Sedgwick
📜 EstateDrafts & assesses wills/trusts; a licensed estate attorney signs before execution (UPL)$2.4BTrust & Will · Wealth.com · Vanilla
💡 PatentPrior-art + patentability; a USPTO-registered practitioner signs before filing (37 CFR 11 / candor)$5.1BSolve Intelligence · IPRally · DeepIP

→ All autopilots · run /flow <vertical> to see any flow in your terminal

Each autopilot keeps a human on the judgment calls — a certified coder, a licensed attorney, a controller, a credentialed preparer. The autopilot does the volume; the human owns the call that carries liability. 22 live connectors run on real data across the 25 verticals — FHIR, ICD-10 (NLM), NCCI/MUE, X12 837P, DocuSign, Plaid, OFAC, staged-rollout, a US federal tax engine, plus medical-necessity criteria, claims fraud-scoring, mortgage AUS (DU/LPA), OIG/SAM exclusion screening, FMCSA carrier-vetting, Reg-F/TCPA outreach guardrails, IOC threat-intel, and FinCEN SAR generation. They're keyless by default (public source or deterministic real generation) and POST to the real provider the moment you add credentials.

Operate: the console where a human signs

→ The buyer-facing story of this surface: greatcto.systems/operate

great-cto board opens the operator console at http://localhost:3141/autopilot.html — the Operate-mode surface where the autopilot's work lands for a named human. Every case the autopilot can't auto-clear waits in the inbox with its AI recommendation, confidence, evidence and an SLA clock; signing executes the irreversible write, rejecting stops it. Nothing irreversible runs without a signature.

Two surfaces, one engine. The builder board (you) and the operator console (your signers) are separate faces of the same server:

great-cto board                    # builder: kanban, agents, pipeline — localhost only
great-cto console --port 8788      # operator console ONLY — the dev board does not exist here
great-cto console --bind 0.0.0.0   # hostable (tunnel / console.client.com); invite-only entry

On the console surface there is no local-admin fallback — without a valid invite link the operator sees an "ask your administrator" screen. An invite token never opens builder routes (any mode), and an invited operator acts strictly inside their own tenant. Invite links expire after 7 days (GREAT_CTO_INVITE_TTL_DAYS, 0 = never).

Quick start: onboarding your signers (one npx great-cto install covers both surfaces):

  1. Admin opens the console → Team panel → picks the operator's role (underwriter, BSA officer, certified coder… — 26 roles) and tenant → Create invite → copies the link (or emails it). The link expires in 7 days.
  2. Operator opens the link — and is in: role and workspace are baked into the token and can't be changed from the browser. They see only the console (tenant's brand if set), only their tenant's cases.
  3. Work: cases land in the inbox — started by hand, or pushed by a source system via POST /api/autopilot/ingest (HMAC-signed with GREAT_CTO_INGEST_SECRET). The operator reads the AI recommendation + evidence and signs or rejects; every decision is recorded with date, time, evidence and confidence in the audit trail.
  4. Revoke: Team → Revoke kills the token instantly.

When binding beyond loopback (--bind), operators are protected by invite tokens; put your reverse-proxy auth in front for anything admin-grade — full hosted auth ships with the console package split (P5 in docs/plans/PLAN-ui-split.md).

Under the hood (for the CTO who runs it)

→ The builder-facing story of this surface: greatcto.systems/build

Each autopilot is built and operated by a gated pipeline of specialist agents — architect, 12-angle reviewer, QA, security officer, devops — tuned to your stack and jurisdiction. You make two decisions per feature; everything else runs automatically. Every vertical ships with its own domain compliance reviewer — False Claims Act & NCCI for coding, OFAC & BSA for AML, FDCPA & Reg F for collections, 21 CFR Part 11 for trials, ECOA for lending, ALTA for title, FMCSA for freight — that blocks an unsafe design before it ever runs. Each reviewer is held to an adversarial golden test set in CI before release. The reviewers, signed human gates, audit trail, and live connectors are the trust layer that makes it safe to let the autopilot run.

Recommended companion MCP: Serena (semantic code navigation). On large codebases the code-writing agents (senior-dev, coder) burn context grepping and reading whole files. The Serena MCP gives them symbol-level navigation (find-symbol, references, structure) instead:

claude mcp add serena -- uvx --from git+https://github.com/oraios/serena \
  serena start-mcp-server --context ide-assistant --project "$(pwd)"

Optional — everything works without it; with it, implementation tasks on big repos use noticeably less context per edit.

The permission is never the wound. Every flow step is tagged reversible or not; the runtime refuses to execute an irreversible action autonomously — money moves, claim submission, e-signing, fleet changes and tax filing run only after a named human signs the checkpoint. Each autopilot also declares an accountable owner — one person answers for what it does. flow-runner.mjs <vertical> --validate enforces the invariant; all twenty-five autopilots ship green.

By the numbers

One regulated feature, end to end (voice-AI compliance pack, traced)1h 26m · $3.40 LLM vs ~$42K / ~6 weeks traditional
An earlier CLI-feature run, same pipeline$2.39 LLM vs ~$5,460 human-equivalent; security caught 2 defects QA had passed
Monthly cost (20 pipeline runs)~$34
Autopilot verticals25 (healthcare · finance · legal · ops — each with a human gate)
Specialist agents83
Archetypes auto-detected26
Jurisdictions12 (GDPR · HIPAA · PCI-DSS · SOX · and more)

→ Full trace with all artefacts

How it works

npx great-cto init — scans your stack and README, detects jurisdiction (GDPR? HIPAA? PCI?), writes .great_cto/FLOW.md with the exact agents, gates, and compliance frameworks for your project.

/start "describe the feature" — critics review the architecture and spec before any code is written. You review the plan at gate:plan.

Agents run automatically — senior-dev implements with TDD, 12-angle review, QA, security, devops. You approve ship at gate:ship.

Three projects — three different pipelines

Same command. Output depends on what you're building and where it runs:

Fintech startup · EUHealthcare portal · USCLI tool
Specialist agentspci-reviewer · gdpr-reviewer · regulated-reviewerfda-reviewer · healthcare-reviewer · security-officercli-reviewer
Human gatesgate:gdpr-dpia · gate:plan · gate:shipgate:clinical-validation · gate:plan · gate:shipgate:plan
ComplianceGDPR · PCI-DSS · SOXHIPAA · HITECH—
Cost / cycle~$8–18~$8–18~$0.5–3

→ Try the interactive picker: greatcto.systems/#flow-picker

The dashboard you'll actually check

great-cto board opens at http://localhost:3141 — Kanban with realtime SSE, per-agent cost tile, pipeline status, 30-day LLM spend vs human-equivalent baseline.

One builder, many operators. Build is for the one-person engineering org — an indie hacker, solo founder, or technical CTO running the pipeline on Claude Code or OpenAI Codex. Operate is for everyone who signs the work: licensed adjusters, attorneys, controllers, compliance leads — invited into the operator console with scoped, tenant-isolated links. One engineer builds the autopilot; the whole back office runs on it. Not for multi-dev engineering teams — see FAQ.

Install

npx great-cto init

Restart your AI host after init. Requires: Node 18.17+ and one of:

HostInstall flagStatus
Claude Code(default)✅ full support
OpenAI Codex--host codex✅ hooks + MCP + agents
# Claude Code (default)
npx great-cto init

# OpenAI Codex Desktop / CLI
npx great-cto init --host codex

Superpowers and Beads companion plugins install automatically — no manual setup needed.


Two decisions per feature

🟡 gate:plan   ←  you decide here (architecture + tasks + cost)
   ↓
🤖 senior-dev → 12-angle review → qa-engineer → security-officer → devops
   ↓
🟢 gate:ship   ←  you decide here (PR ready, security signed off)

Architects, planners, reviewers, QA, security, DevOps run automatically between those two human checkpoints. Memory persists between sessions: every gate verdict appends to ~/.great_cto/decisions.md, every retrospective appends to per-project lessons.md, and /crystallize promotes high-impact patterns to a global library agents query before re-solving.

Critics before the plan

The most expensive bugs aren't in the code — they're in decisions made before coding starts. Three critic agents run before the Plan stage, at the three positions where a mistake costs the most:

CriticCatches
Architecture criticCoupling that rules out multi-tenancy later · "obvious" O(n²) on real-scale data · circular dependencies between bounded contexts
Spec critic"We solved the wrong problem" — the worst class of bug, because no unit test will catch it · misaligned acceptance criteria · scope that was never agreed on
Schema criticNOT NULL without a default on a 50M-row table (deadlock in 10min after deploy) · missing CONCURRENTLY on index creation · irreversible migrations with no rollback path

Previously critics only activated starting from Plan. Now the pipeline catches architectural and spec-level mistakes before implementation begins — when reverting costs hours, not days.

How great_cto compares

great_ctoDevinClaude Code (alone)
Open source✅ MIT❌ closed❌ closed plugin model
Self-host✅ runs locally❌ Cognition cloud✅
Host✅ Claude Code + Codex❌ Cognition cloud✅ Claude Code
BYOK / multi-model✅ Claude Code · Codex❌ proprietary❌ Anthropic only
Specialist agents83 (architect · PM · 12-angle review · QA · security · devops · reviewers across archetypes, packs & jurisdictions)1 generalist1 generalist
SDLC orchestrationarchitect → plan → impl → review → QA → security → devopsone-shot autonomyedit loop
Human gates✅ 2 per feature (plan + ship)❌ none❌
Memory across sessions✅ decisions.md + lessons.md + crystallize⚠️ thread only⚠️ thread only
Cost tracking✅ per-agent + 30d history + savings_x❌❌
Compliance frameworks✅ 45+ (PCI · HIPAA · SOX · GDPR · CCPA · DPDPA · EU AI Act · FDA SaMD · COPPA · FERPA · FedRAMP · NAIC · …)❌❌
Pricingfree (you pay your LLM provider)$500/mo$20/mo
Setupnpx great-cto initsign upinstall CLI

great_cto is not another coding-agent loop — it's the orchestration layer above the coding agent you already use. Think "specialist team that reviews and gates the work" rather than "another assistant that types code."

Jurisdiction detection

npx great-cto init scans three signal sources — README keywords, infra region strings (Terraform, .env AWS_REGION=, docker-compose TZ=), and package.json homepage TLD — and auto-detects which of 12 jurisdictions apply:

JurisdictionSignals (README + infra)FrameworksReviewer
eugdpr · eu users · nis2 · eu ai act · eu-west-* · .de TLDGDPR · EU AI Act · NIS2 · ePrivacygdpr-reviewer
us-caccpa · cpra · california residents · do not sellCCPA / CPRAus-privacy-reviewer
ukuk gdpr · information commissioner · dpa 2018UK GDPR · DPA 2018gdpr-reviewer
indpdpa · india users · rbi data localisationDPDPA 2023 · RBIdpdpa-reviewer
brlgpd · anpd · brazil usersLGPDgdpr-reviewer
auprivacy act 1988 · oaic · notifiable data breachPrivacy Act 1988 · CDRus-privacy-reviewer
sgpdpa · pdpc · mas guidelines · singpassPDPA · MAS TRMus-privacy-reviewer
capipeda · quebec law 25 · casl · canadian users · ca-central-*PIPEDA · Quebec Law 25 · CASL · OSFI B-10us-privacy-reviewer
jpappi · japan users · my number · ap-northeast-1 · japaneastAPPI 2022 · PPC Guidelines · FISCus-privacy-reviewer
cnpipl · mlps · china users · cn-north-* · cn-east-*PIPL 2021 · DSL 2021 · MLPS 2.0 · CBDTgdpr-reviewer
krpipa korea · isms-p · kisa · korea users · ap-northeast-2PIPA · ISMS-P · FSC regulationsus-privacy-reviewer
usftc · us users · virginia cdpa · texas tdpsaFTC Act · US state privacy lawsus-privacy-reviewer

Word-boundary matching prevents false positives ("india" doesn't match "indiana"). Detected jurisdiction is written to PROJECT.md as jurisdiction: [eu, us-ca] and gates the appropriate reviewer on every feature. Override manually:

jurisdiction: [eu, us-ca]

Three commands you use every day

/start "build a refund endpoint with PCI-DSS scoping"
# → architect → enterprise-saas-reviewer (PCI-DSS auto-loaded)
# → pm → 5 Beads tasks → gate:plan (you approve)
# → senior-dev → 12-angle review → qa → security-officer
# → gate:ship (you approve) → devops → deployed

/inbox
# Pending gates · P0 incidents · blocked tasks · stale in-progress

/digest
# Weekly DORA + delta vs last week + cost-per-feature roll-up

Plus: /audit (existing-codebase scan), /cost (LLM router savings), /sec (security umbrella), /oncall, /release, /rfc. Full list: ~/.claude/commands/ after install.

Cost

~$34/month for a typical solo-CTO project — 20 pipeline runs/month, indicative.
PipelineCost/runRuns/moTotal
quick (config / typo)$0.1010$1
quick (new endpoint)$16$6
standard (feature)$53$15
deep (cross-cutting)$121$12
~$34

Pay your own Anthropic API tokens. No per-seat fee. No SaaS lock-in. Routine triage auto-routes to Kimi K2 (Sonnet-equivalent at ~5× lower cost) → 60–80% reduction on log clustering.

26 archetypes auto-detected

Each archetype activates its own specialist agents and compliance checklists. Top 7:

ArchetypeTierSpecialist agentsCompliance
enterprise-saasdeepenterprise-saas-reviewersoc2-type-2 · iso27001 · gdpr · ccpa
agent-productdeepai-prompt-architect · ai-eval · ai-securityeu-ai-act · owasp-llm-top-10
fintechdeeppci · regulatedpci-dss · sox · kyc-aml · gdpr · dora
mlopsdeepmlops-reviewer · ai-evaleu-ai-act · nist-ai-rmf · iso42001
librarybaselinelibrary-revieweropenssf · sbom
cli-toolbaselinecli-reviewer—
mobile-appstandardmobile-store-reviewerstore-policy · gdpr
defense-govcondeepcmmc-reviewer · gov-reviewercmmc-2.0 · nist-800-171 · dfars · itar · section-889

Full table (26 archetypes) + how detection works: docs/ARCHETYPES.md.

Deep US coverage — beyond GDPR/PCI/HIPAA, great_cto now reviews against SEC cyber-disclosure (8-K Item 1.05), CMMC 2.0 / NIST 800-171 for defense contractors, US AI governance (NIST AI RMF · Colorado SB 205 · Utah/Texas AI), web-tracking litigation (VPPA · CIPA · Washington MHMDA), and HMDA / SR 11-7 model risk for lending.

14 domain packs — overlay reviewers

Domain packs ride on top of archetypes. Auto-attached when CLI detects pack-specific signals (deps, README terms). Each pack adds its own reviewer(s), threat-model template, EVAL suite, and human gates — independent of base archetype.

CategoryPacks
AI verticalsvoice-pack · clinical-pack · hr-ai-pack · drug-discovery-pack
Digital healthdigital-health-pack (wearable telemetry · mental-health AI · nutrition AI · physician HITL)
Fintech / regulatedlending-pack · em-fintech-pack
High-complianceclinical-trials-pack · climate-pack
Engineeringapi-platform-pack · robotics-pack
US marketsec-cyber-pack (SEC 8-K disclosure) · adtech-privacy-pack (VPPA · CIPA · MHMDA) · us-ai-pack (NIST AI RMF · Colorado SB 205)

→ 28 human-gate types + 53 reference EVAL suites + 15 TM templates. Browse all 14 packs with 4-layer journey visualization (archetype → pack → reviewer → gate): greatcto.systems/packs.html.

One real run, fully traced

The canonical receipt: a voice-AI compliance pack (TCPA screening, STIR/SHAKEN, state recording-consent) shipped through the full pipeline in 1h 26m wall-clock for $3.40 in LLM cost — architect → threat model → implementation → 5 reviewers → human gates → merged PR. The traditional path for the same regulated feature: ~170 hours and ~$42K. Every stage timestamped, every artifact links to a public GitHub PR.

An earlier run on a Python CLI feature ($2.39 vs ~$5,460 human-equivalent) showed the review model working: security caught two real defects QA had passed (list(stream_csv()) defeated streaming → 14.5 MB peak RSS on 13 MB input).

Full trace + artefacts: greatcto.systems/proof · raw: docs/qa/runs/2026-05-09/E2E-CLI-PIPELINE.md.

CI integration

Drop into any GitHub Actions workflow:

- run: npx great-cto@latest ci ./ --sarif results.sarif
- uses: github/codeql-action/upload-sarif@v3
  if: always()
  with: { sarif_file: results.sarif }

great-cto ci auto-detects $GITHUB_ACTIONS and emits ::error file=...,line=N:: annotations inline on PR diffs. Exit codes: 0 clean / 1 findings / 2 setup error.

Test pyramid

Layered test suite — structural + state-machine tier runs in <2 min for $0 (node --test tests/*.test.mjs); real-LLM tier (26 archetypes × 4-8 stages + 14 packs + 13 reviewers) runs on-demand via OpenRouter for ~$5–10. Full breakdown: docs/testing/.

MCP

Native MCP server — 7 tools callable from Claude Desktop, Codex, or any MCP host. Local (no board needed): detect_archetype · estimate_cost · query_decisions. Board-backed: project_status · cost_summary · pipeline_stages · recent_verdicts.

{ "mcpServers": { "great-cto": { "command": "npx", "args": ["-y", "great-cto@latest", "mcp"] } } }

Full setup + internal MCPs (Grafana, LLM router, Beads): docs/MCP.md.

Email alerts (zero-setup)

Five things that need you to act in <2h get emailed automatically — even when you're away from the board:

TriggerWhen
🚨 P0 incidentA P0 task opens in any project
⏸️ Gate stale > 2hA gate:ship is waiting on you for hours
🛡️ Security BLOCKEDsecurity-officer rejected a merge
💸 Budget alertMonthly LLM spend crosses 80% / 100% of budget
📊 Weekly digestFriday 09:00 — shipped, spent, savings, QA

Setup: board → Notifications tab → enter email → enter the 6-digit code we send → pick triggers. No Resend signup, no API keys — delivery routed through greatcto.systems/notify (free, 100 emails/24h per verified email).

Limitations & non-goals

  • Not for multi-dev engineering teams — one builder is the product; 2+ engineers sharing the pipeline have outgrown it. Operators are unlimited — invite signers and compliance leads to the console freely.
  • Not a replacement for senior engineers — codifies process; doesn't make architectural judgement calls without one.
  • Not a CI/CD system — gates run locally / in-session. You still need GitHub Actions for actual merge.
  • Not certification-audited — PCI/HIPAA/SOC2 archetype scaffolds are starting points, not certifications.
  • Not deterministic — LLM-generated outputs. Every gate verdict should be sanity-checked.

FAQ (top 5)

Is my source code used to train models? No. Claude API zero-retention by default for paying customers. great_cto adds nothing.

How do you keep token costs down? Haiku-by-default + Kimi K2 router for triage (60–80% savings) + cost-guard hook.

Can I disable hooks? Every hook honors GREAT_CTO_DISABLE_<NAME>=1. Per-file secret-scan opt-out: // great_cto:allow-secrets.

What if I'm not solo? The engineering side is built for one person — if you have 2+ engineers who need shared builder boards, you've outgrown it. The operating side is multi-user by design: invite as many signers and compliance leads to the operator console as your back office needs (scoped invite links, per-tenant isolation).

Full FAQ: docs/FAQ.md.

Documentation

📚 Full documentation hub → — organized by Diátaxis: Getting Started · How-to guides · Agents & Commands reference · Architecture · FAQ.

Architecture

The plugin runs inside Claude Code (or any MCP-capable host); 83 agents are markdown specs; tasks live in Beads (dolt, git-native); memory is plain markdown (no vector store). Diagram + stack table: docs/ARCHITECTURE.md.

What's new

v2.40–v2.62 (June 2026) — The autopilot pivot: GreatCTO becomes AI autopilots for business — 25 service-autopilot verticals, each a flow with a measured quality scorecard, an accountable owner, and the runtime invariant that an irreversible action never executes without a human signature. 22 live connectors run every vertical on real data. Story: We pivoted →

v2.46–v2.63 (June 2026) — The operator console: durable runs pause at the human gate and wait in an inbox for a named licensed human; signing executes the write. Role-based access, scoped invites, AI-drafted determinations with evidence, QA sampling, SLA clocks, Ops tab (metering · connector health · dead-letter requeue), WCAG 2.2 AA, light/dark. Story: The operator console →

v2.37–v2.65 (June 2026) — Under the hood: the dev board becomes a pult — approving a gate can spawn a live-streamed agent run; prompt self-improvement gated on held-out evals (SIA-inspired); $0 context compression (CI log 31,475 → 155 chars with the FATAL preserved); Fable 5 support. Story: June under the hood →

Full changelog →

Roadmap

  • Hosted operator console — one-command tunnel + custom domain for great-cto console, so signers never need localhost
  • Vertical depth over breadth — push the measured quality scorecard ≥95 on the top-5 autopilots before adding new ones
  • SOC 2 evidence pack — export the audit trail + gate history in auditor-ready format
  • Multi-model verification — independent second-model review on irreversible-action gates

Vote on the next feature →

Author

avelikiy — CTO building AI-native trading and fintech platforms (0→1, 1→N). great_cto is the result of automating my own loops, one agent at a time. Every rule appeared in response to a real problem in a real production system.

Community

ChannelWhat
🐛 IssuesBugs, feature requests, archetype proposals
💡 DiscussionsQuestions, patterns, show-and-tell
📝 BlogReceipts, cost breakdowns, architecture deep-dives
🔒 SECURITY.mdResponsible disclosure

Contributing & License

Pull requests welcome — see CONTRIBUTING.md. Good first issues: good-first-issue.

MIT — see LICENSE.

If great_cto saved you time, please star the repo — it helps other solo CTOs find it.

Star History Chart


Built by @avelikiy Stop being the only person who can ship.

Ver o original no GitHub

Bom para

Relacionados