Concept Exploration: gblock-party (Final)

Updated 2026-05-25 • Answers integrated • Previous version

Resolved Decisions

Q1: Personal — Value Wedge
Development priority for the personal version
Agent Board + Terminal first
Impact: VS Code, Browser Preview, and File Explorer are deferred. Existing tools (code-server, direct tunnel) cover those needs until orchestration UX is proven.
Q2: Personal — Constraints
Access layer selection
Both in parallel — Tailscale for terminal, Cloudflare Tunnel for web dashboard/PWA
User note: "Tailscale for terminal, Cloudflare Tunnel for Website"
Impact: Confirmed. Tailscale provides direct, low-latency mesh VPN access for terminal/SSH sessions. Cloudflare Tunnel provides authenticated web-facing URLs for the Agent Board PWA and preview ports. They complement each other — Tailscale for dev-level access, Cloudflare for the web control surface.
Q3: Personal — Unknowns
CLI integration investigation approach
Technical spike before any UI work
User note: "I'm worried about Claude Code terms of service, so we might just focus on codex cli for now."
Impact: Major pivot. Focus on Codex CLI first for programmatic control. Claude Code integration deferred pending ToS review. Codex CLI has codex exec --json for JSONL event streaming, which is significantly more automation-friendly. See Codex CLI Reference below.
Q4: SaaS — Beneficiary
ICP priority segment
Power users / solo founders
User note: "we'll leave this to the ICP to investigate and confirm"
Impact: Starting hypothesis for ICP analysis. ICP will validate or redirect.
Q5: SaaS — Unknowns
Pricing confidence
Genuinely unknown
User note: "Let the monetization skill investigate and confirm that"
Impact: Removed all pricing anchors from concept briefs. Pricing deferred entirely to monetization research. Do not reference $29-79 or $199-499 ranges as starting points.
Q6: SaaS — Next Steps
Post-brief sequencing
Business discovery first
Impact: /pack install business-discovery/icp → then technical spike on Codex CLI. Market validation before engineering investment.

Brief 1 — Personal

Summary

A persistent Hetzner-powered remote development machine that runs multiple Codex CLI sessions in parallel, controlled from browser or phone. Five integrated surfaces — Agent Board, Terminal, VS Code, Browser Preview, and File Explorer — provide a complete AI coding workstation with zero public inbound access.

This is the personal dogfood environment. Agent Board + Terminal are the priority surfaces. VS Code, Preview, and File Explorer are deferred to existing tools.

Problem Hypothesis

Running AI coding agents locally is constrained by:

Beneficiary Hypothesis

The sole beneficiary is you — a solo developer who wants to run 5-12 parallel AI coding sessions across multiple repos, monitor them from Mac and iPhone, and maintain a locked-down security posture.

Value Wedge Updated

Human-in-the-loop multi-agent orchestration from any device.

Development Priority New

Agent Board + Terminal first. These are the two most critical surfaces. VS Code, Browser Preview, and File Explorer are deferred — existing tools (code-server, direct Tailscale tunnel) cover those needs until the core orchestration UX is proven.

Constraints Updated

ConstraintDetailSource
ProviderHetzner Cloud USDecided
Default machineCCX33 (8 vCPU, 32 GB RAM)Decided
Burst machineCCX43 (16 vCPU, 64 GB RAM)Decided
Scaling ruleDo not expand root disk on scale-upDecided
OSUbuntu LTSDecided
Network postureZero public inbound portsDecided
Access layerBoth: Tailscale for direct terminal/SSH, Cloudflare Tunnel for web dashboard/PWAResolved Q2
Agent CLIsCodex CLI first (better programmatic API via exec --json); Claude Code deferred pending ToS reviewResolved Q3
Session managementtmux-basedDecided
Browser IDEcode-server or OpenVSCode Server (deferred surface)Decided
Code structureMonorepo shared with commercial SaaSDecided

Non-Goals

Assumptions & Unknowns Updated

Confirmed Assumptions

Unknowns Requiring Validation

#UnknownRiskNotes
1Codex CLI programmatic controlCriticalTechnical spike before any UI. Codex has exec --json (JSONL events) and -o FILE. Spike should validate whether these provide sufficient control for the Agent Board.
2Permission prompt interceptionCriticalCodex has --dangerously-bypass-approvals-and-sandbox and -s MODE sandbox policies. May provide a path for web-based approvals.
3Session state observationHighCodex --json JSONL output may stream status events. Needs spike validation.
3.5Claude Code ToS complianceHighPotential terms-of-service concerns with programmatic control. Focus on Codex first; revisit after ToS review.
4Output streamingMediumHow to stream agent output to web UI in real time.
5Resource ceilingLowWhether CCX33 handles 10+ parallel sessions. Testable empirically.

Codex CLI Reference New

Key Codex CLI capabilities relevant to the Agent Board:

codex exec "prompt" — Run a prompt non-interactively (alias: codex e)
--json — Print events to stdout as JSONL (state observation, output streaming)
-o, --output-last-message FILE — Write agent's final message to a file
--output-schema FILE — Constrain final response shape with JSON Schema
-s, --sandbox MODE — Sandbox policy: read-only, workspace-write, danger-full-access
--dangerously-bypass-approvals-and-sandbox — Skip prompts (for externally sandboxed envs)
--ephemeral — Don't persist session files
--color never — Disable color output (for piping)
-m, --model MODEL — Select model
-i, --image FILE — Attach image(s)
-p PROFILE — Select config profile (NOT equivalent to Claude Code's -p)

Note: Codex's -p selects a configuration profile from config.toml, not print/prompt mode. The equivalent of Claude Code's -p is codex exec. Stdin is also supported — if no prompt argument is given, instructions are read from stdin.

Next Steps Updated

Resolved sequencing: Business discovery first, then technical spike.

  1. /pack install business-discovery — install research skills for commercial concept.
  2. /icp — validate beneficiary hypotheses (starting with power users / solo founders).
  3. Technical spike on codex exec --json — prove feasibility of programmatic control before any UI work.
  4. /feature-interview on Agent Board + Terminal once CLI strategy is confirmed.

Brief 2 — Commercial SaaS

Summary

A browser/mobile SaaS control plane for AI coding agents on gblock-party's managed cloud infrastructure. Users connect their preferred front-end tool (T3 Code, Multica, Claude Code CLI) or use the native Agent Board.

All infrastructure is managed by gblock-party on Hetzner.

Builds on the personal workstation codebase via a shared monorepo.

Problem Hypothesis

Developers increasingly use AI coding agents but face these problems at scale:

Beneficiary Hypothesis Updated

Primary ICP starting hypothesis: Power users / solo founders

To be validated by ICP analysis. Other segments for later investigation:

Secondary (managed IaaS tier)

Value Wedge

Human-in-the-loop multi-agent orchestration from browser and mobile.

vs.Gblock Party advantage
Local CLI usageRemote access, mobile, multi-agent visibility, session persistence
Browser IDEs (Gitpod, Codespaces)Agent orchestration focus, not just "code in browser"
T3 CodeBetter mobile UX, zero-inbound security, session management, diff review, approval workflow
Self-managed VPS + tmuxTurnkey control plane, no sysadmin, Agent Board UX

Constraints Updated

ConstraintDetailSource
Build orderPersonal workstation built and dogfooded firstDecided
Code structureMonorepo shared with personal versionDecided
Commercial modelHybrid SaaS + optional managed IaaSDecided
Host OSUbuntu LTS (initial target)Decided
Network postureHost-initiated outbound onlyDecided
Agent CLIsCodex CLI first; Claude Code deferred pending ToS review; OpenCode laterResolved Q3
Managed infra providerHetzner (for optional IaaS tier)Decided
No GitHub ActionsProject does not use GitHub ActionsDecided

Assumptions & Unknowns Updated

Confirmed Assumptions

Unknowns Requiring Validation

#UnknownRiskCategoryNotes
1Codex CLI programmatic controlCriticalFeasibilitySpike on codex exec --json. Must be resolved during personal dogfood.
2Permission prompt interceptionCriticalFeasibilityCan prompts be routed through web UI?
3Session state observationHighFeasibilityReal-time status from Codex JSONL output.
4Daemon-to-SaaS tunnelHighArchitectureTechnology choice undecided.
5Willingness to payHighMarketPricing genuinely unknown. Defer to monetization research — no anchoring on preliminary ranges.
6Data custody trustMediumMarketTrust barrier for tunneling through SaaS.
7Daemon reliabilityMediumEngineeringAuto-updates, crash recovery on gblock-party's managed infrastructure.
8Multi-agent CLI compatibilityMediumFeasibilityWhether future agents have similar interfaces.

ICP Readiness

Ready for ICP once /pack install business-discovery is run.

ICP Inputs

Assumptions to Test First

  1. Codex CLI programmatic control feasibility (technical spike).
  2. Whether target users currently run multiple AI agents in parallel.
  3. Whether zero-inbound security posture matters or is table stakes.

Next Steps Updated

Resolved sequencing: Business discovery first.

  1. /pack install business-discovery — install ICP, competitive analysis, value prop, positioning, lean canvas.
  2. /icp — starting with power users / solo founders hypothesis.
  3. Monetization research for pricing (no anchoring on preliminary ranges).
  4. Technical spike on Codex CLI integration (after business discovery).