Journey Map — gblock-party

Full user and customer lifecycle for the Agent Conductor across the two-track model (Track A: BYO-client with Agent Board as dashboard · Track B: Agent Board as primary UI). Product: gblock-party · Date: 2026-05-31 · Based on: research/icp.md, research/competitive-analysis.md, concept briefs.

Summary

The Agent Conductor (primary ICP: solo developer / indie hacker running 2-10+ AI coding agents) follows a lifecycle driven by infrastructure scaling triggers and mobile access needs across two tracks: Track A (BYO-client) [BYO-client connectivity needs validation] connects an existing front-end tool (T3 Code, Multica, Claude Code CLI) to gblock-party's managed infrastructure — the user's own client remains their primary coding UI, and Agent Board serves as an infrastructure dashboard for session monitoring, resource status, and mobile check-ins; Track B (Agent Board) is for users without an existing tool preference who adopt gblock-party's native Agent Board as their primary UI. gblock-party manages the server infrastructure in both tracks. Five core user journeys map the task-level experience; the customer lifecycle maps the relationship from trigger through advocacy and churn.

The product wins or loses at five critical moments: first setup (5-minute bar), first laptop close (persistence proof), first mobile approval (phone proof), first review bottleneck (scale proof), and first bill (value justification vs. DIY alternatives).

The aha moment, in users' own words: "I started 5 agents, closed my laptop, had dinner, checked from my phone — 3 PRs were ready for review."

Two-Track Framing

Is the two-track journey model (Track A BYO-client + Track B Agent Board, converging on managed infrastructure) the right framing for the whole map?

Yes — two-track (BYO-client + Agent Board) converging on managed infra is the right frame. No — collapse to a single track; the BYO-client vs Agent Board split overcomplicates the map. Needs more tracks/personas beyond solo Agent Conductor (e.g. team lead) before this framing holds. Other / None of the above Need clarification

User Journeys

Primary persona: The Agent Conductor — solo developer using 2-4 AI coding tools simultaneously. Enters via Track A (BYO-client: existing front-end tool, Agent Board as infrastructure dashboard) or Track B (Agent Board as primary UI for users without an existing tool preference).

Journey 1: Scale Past the Laptop

Trigger: Running 3-5 agents locally, laptop CPU/RAM maxes out

Track A: BYO-client

Already uses a front-end tool (T3 Code, Multica, Claude Code CLI)
Discovers gblock-party (blog / HN / Reddit)
Signs up for gblock-party, gets API credentials
Connects existing tool to gblock-party's managed infrastructure
Agents now run on persistent cloud servers

Track B: Agent Board

No existing tool preference — open to adopting Agent Board as primary UI — discovers gblock-party (blog / HN / Reddit)
Signs up, clicks "Create Environment"
gblock-party provisions managed environment
Agent Board connects to managed infrastructure

Both tracks converge on managed infrastructure

Track A: user continues working in their own client; Agent Board available as infrastructure dashboard (session status, resource monitoring, mobile check-ins)
Track B: Agent Board loads as primary UI — connected host visible, start first agent session
Both: see all agents in one unified view via Agent Board

"Is this meaningfully better than my current DIY setup?"

Sessions persist across devices, mobile monitoring works; Track A users keep their preferred workflow, Track B users get a turnkey UI

Setup friction too high, or orchestration adds latency to agent execution (T3 Code 4x slowdown warning)

Journey 2: Check from Phone

Trigger: Away from desk, agent needs approval or long-running task finishes

Receive push notification on phone
Open PWA — see agent status at a glance
Review approval context (action, risk level, affected files)
Approve or reject with one tap
Optionally review diff before agent pushes
Resume other activity

"Can I actually do useful work from my phone in under 30 seconds?"

Push notification → one-tap approve → back to life in <30 seconds

Mobile UI too cramped for diffs, approval context insufficient to make safe decisions

Journey 3: Close Laptop, Keep Working

Trigger: End of work session, agents mid-task

Close laptop — agents continue running on gblock-party's managed infrastructure
Agents complete tasks, queue approvals if needed
Check next morning from any device (laptop, phone, tablet)
Review "what happened while I was away" summary
Review completed diffs, approve pushes

"Did I lose any state? Can I see exactly what happened?"

All sessions exactly where they were, output captured, diffs ready for review, session recap available

Session state corrupted, output lost, no clear summary of what changed overnight

Journey 4: Parallel Sprint

Trigger: Multiple features/bugs to ship, wants to parallelize across agents

Start 5-8 agents across repos with specific prompts/specs
Monitor progress via Agent Board — per-agent status cards
Handle approval prompts as they arrive (desktop or mobile)
Git worktree isolation prevents merge conflicts
Batch-review diffs across agents
Ship PRs for completed work

"Can I track what each agent is doing without context-switching between terminal windows?"

Board shows per-agent status, worktree isolation works, batch review flow keeps velocity high

Merge conflicts from agents on shared files, review bottleneck overwhelms (can only review so fast), 3-5 agent sweet spot exceeded

Journey 5: First-Time Setup

Trigger: Discovered gblock-party, ready to try it

Track A: BYO-client

Sign up free (self-serve, no procurement)
Copy API credentials from dashboard
Configure existing front-end tool to connect to gblock-party
Tool connects — sessions now run on persistent infrastructure; user continues coding in their own client

Track B: Agent Board

Sign up free (self-serve, no procurement)
Click "Create Environment" in dashboard
gblock-party provisions managed environment
Agent Board loads as primary UI showing managed environment

Both tracks converge on managed infrastructure

Track A: start first agent session from own client; open Agent Board dashboard for monitoring
Track B: start first agent session from Agent Board
Both: bookmark PWA on phone for mobile access

"How long until I see value?" (target: under 5 minutes)

Signup → connect → first agent running in under 5 minutes, zero config

Track A: API credential setup confusing, tool integration docs missing, connection fails silently. Track B: provisioning takes too long, unclear status during wait.

Journey Coverage

Five journeys are mapped (Scale Past Laptop, Check from Phone, Close Laptop Keep Working, Parallel Sprint, First-Time Setup). Are these the right user journeys, and is anything missing?

Yes — these five cover the core workflows for the Agent Conductor. Missing a journey (e.g. a team-collaboration or billing/upgrade journey). Right journeys, wrong priority order. Other / None of the above Need clarification

Customer Lifecycle

Trigger

Scaling past laptop

Discovery

Blog / HN / Reddit

Evaluation

Free signup, try

Onboarding

Connect tool / board

Aha

Laptop close + phone

Conversion

Free → paid

Retention

Daily Agent Board

Expansion

Solo → team

Stage	What Happens	Key Metric
Trigger	Laptop maxes out at 3-5 agents, or first mobile access need (away from desk, agent needs approval). Searches for remote agent hosting solutions.	—
Discovery	Finds gblock-party via blog post, Show HN, Reddit thread, or tutorial search ("run Claude Code remotely", "AI agent orchestration"). Peer recommendation in Discord/Slack.	Traffic source, CTR
Evaluation	Free signup. Track A: connects existing front-end tool to managed infra, compares to DIY setup; Agent Board available as dashboard but not the primary coding UI. Track B: creates managed environment, evaluates Agent Board as primary UI. Solo decision in minutes to hours. Checks GitHub stars, open-source signal. Will not pay before trying.	Signup → install rate
Onboarding	Track A: configures existing tool with API credentials, connects to managed infra; user stays in their own client for coding, Agent Board available as infrastructure dashboard. Track B: click "Create Environment", gblock-party provisions managed environment, Agent Board loads as primary UI. First agent session started. PWA bookmarked on phone.	Time to first agent session (<5 min target)
Aha Moment	"I started 5 agents, closed my laptop, had dinner, checked from my phone — 3 PRs were ready for review." Sessions survived laptop closure + successful mobile check.	Sessions surviving laptop close + mobile check-in
Conversion	Free → paid when hitting free tier limits (host count, advanced features) or when value is proven. Self-serve credit card. $9-29/mo flat rate.	Free-to-paid conversion rate
Transaction	Self-serve billing. Flat monthly subscription. No credit-based pricing (avoid Intent's hostile model). Managed devbox add-on available ($49-199/mo).	MRR, ARPU
Retention	Daily use — Track A: own client remains primary coding interface, Agent Board used daily for infrastructure monitoring and mobile check-ins; Track B: Agent Board is primary interface. Session persistence + mobile access create switching cost in both tracks. Value exceeds free DIY alternatives.	DAU/MAU, sessions/day
Expansion	Solo founder hires first engineers → team tier ($29-79/seat). Or adds managed devbox. Natural PLG land-and-expand motion.	NRR, seat expansion
Advocacy	Blog post about workflow, HN comment recommending gblock-party, GitHub star, Discord recommendation to peers.	NPS, referral rate
Churn	First-party tools close the gap (Claude/Codex ship native multi-agent mobile), or workflow changes (stops using AI agents, switches to Cursor cloud).	Churn rate, exit survey
Recovery	Feature announcement re-engagement ("we shipped X you asked for"). Lifecycle email. Community re-engagement via Discord/HN.	Win-back rate

Least-Supported Stage

The lifecycle runs Trigger → Discovery → Evaluation → Onboarding → Aha → Conversion → Transaction → Retention → Expansion → Advocacy → Churn → Recovery. Which stage is least supported by evidence and should be flagged for validation first?

Conversion — the free-to-paid trigger and $9-29/mo pricing is provisional. Expansion — solo → team tier and seat growth is inferred, not observed. Advocacy — the referral / blog / HN advocacy loop is unproven. Recovery — the win-back / re-engagement path is purely hypothetical. Other / None of the above Need clarification

Critical Moments

The 5 moments where gblock-party wins or loses the user, ranked by impact.

First 5 Minutes (Setup)

If setup doesn't work on the first try, the user goes back to DIY (Track A) or abandons signup (Track B). Track A: connecting an existing tool must be frictionless — API credentials, zero server config. Track B: managed environment must provision in under 60 seconds. The 2026 PLG bar is value in under 60 seconds.

Evidence: ICP research — "Will not pay before trying. Need to see value in under 60 seconds."

First Laptop Close (Persistence Proof)

The first time the user closes their laptop and successfully resumes from another device is the aha moment. If sessions are lost or state is corrupted, the core value proposition fails entirely.

Evidence: "Close the laptop, pick up where I left off" is the #1 stated desire across VPS+agent articles and ICP interviews.

First Mobile Approval (Phone Proof)

The first push notification → approve flow on the phone proves "you don't have to be at your desk." If the mobile UX is clunky or the approval context is insufficient for safe decisions, the user won't trust it.

Evidence: 62% of mobile approvals happen from notification banners without opening the app (competitive analysis — Copilot Remote). 93% auto-approve rate signals approval fatigue.

First Review Bottleneck (Scale Proof)

When running 5+ agents, the user hits the review bottleneck. If the diff review UX doesn't help them review faster than raw git diff, parallel agents feel like overhead, not leverage.

Evidence: "The bottleneck shifted from 'AI is too slow' to 'I can only review so fast.'" Flask creator Armin Ronacher limits parallel agents because he can't review fast enough.

First Bill (Value Justification)

The moment the credit card charges. If the user can't articulate why they're paying $19/mo instead of using free DIY tools, they cancel. The delta must be visceral: "I literally could not do this before."

Evidence: ICP — "Must be under $30/mo to compete with free alternatives." High individual price sensitivity. DIY is free, T3 Code is free, Emdash is free.

Stage Detail Index

Each lifecycle stage has a dedicated skill for deeper mapping. This overview stays high-level; stage-level detail belongs in these focused docs.

/onboarding-map

Onboarding

Install flow, first agent session, PWA setup, time-to-value optimization

/conversion-map

Conversion

Free tier limits, upgrade triggers, pricing page, payment flow, objection handling

/transaction-map

Transaction

Billing, plan changes, managed devbox add-on, refunds, trust signals

/retention-map

Retention

Daily engagement loops, feature stickiness, churn signals, recovery paths

/expansion-map

Expansion

Solo → team, managed devbox upsell, seat growth, land-and-expand

Journey Gaps

Open questions exposed by journey mapping that need resolution before or during product design.

Gap 1: Approval UX Fidelity on Mobile

How much context can you show on a phone screen for a safe approve/reject decision? No competitor has solved this well. This needs UX exploration — what's the minimum context (action type, risk level, affected files, diff preview) that enables confident decisions without a full terminal view?

Gap 2: "What Happened While I Was Away" Summary

After laptop close → resume, what's the re-entry experience? The Agent Board needs a session recap view: completed tasks, pending approvals, errors, diffs ready for review. No competitor offers this — they all assume continuous presence.

Gap 3: Review Bottleneck Tooling

If the bottleneck is review speed (not agent speed), what does a purpose-built multi-agent diff review flow look like? Batch review across agents? AI-assisted review summaries? Priority ranking by risk? This is a potential differentiator beyond "just show diffs."

Gap 4: Churn Defense Against First-Party Encroachment

Claude Code Channels, Copilot Remote Control, and Codex Mobile all shipped in Q2 2026. If any of these evolve to multi-agent orchestration, the cross-provider gap narrows. The BYO-client + persistence + agent-agnostic combination must be strong enough to retain users who could switch to a single-vendor solution.

Highest-Risk Gap

Four gaps surfaced by the journey map. Which needs resolution first, and which is best supported by existing evidence?

Gap 1: Approval UX fidelity on mobile — backed by competitive evidence (62% banner approvals, 93% auto-approve). Gap 2: "What happened while I was away" summary — inferred from the aha-moment narrative, no direct evidence. Gap 3: Review bottleneck tooling — backed by Ronacher / "can only review so fast" evidence. Gap 4: Churn defense against first-party encroachment — backed by competitive-analysis ship dates. Other / None of the above Need clarification

Evidence Matrix

Separates OBSERVED user/customer evidence (from research/icp.md, research/competitive-analysis.md, concept briefs) from INFERRED journey stages with no direct evidence. The Assumption-status column flags which stages are inferred.

Claim	Source / repo evidence	Inference	Confidence	Assumption status	Decision impact
Aha moment is "closed laptop, checked from phone, PRs ready"	ICP interviews / VPS+agent articles; user-voice quote	Direct user voice; restated as aha narrative	High	Observed (user evidence)	Anchors Persistence + Phone proof critical moments
"Close the laptop, pick up where I left off" is the #1 stated desire	VPS+agent articles + ICP interviews	Aggregated across sources, ranked #1	High	Observed (user evidence)	Makes persistence the core value prop
Will not pay before trying; value in under 60 seconds	research/icp.md	Direct ICP quote drives 5-min setup bar	High	Observed (user evidence)	Sets onboarding / First-5-Minutes bar
62% of mobile approvals from notification banners; 93% auto-approve rate	research/competitive-analysis.md (Copilot Remote)	Competitor telemetry; implies approval fatigue	High	Observed (competitor evidence)	Shapes mobile approval UX (Gap 1)
Review is the bottleneck ("can only review so fast")	Armin Ronacher (Flask creator) public statement	Named-practitioner evidence generalized to ICP	High	Observed (practitioner evidence)	Justifies review-tooling differentiator (Gap 3, Critical Moment 4)
T3 Code orchestration risks ~4x slowdown	research/competitive-analysis.md (T3 Code)	Competitor signal applied as failure mode	Medium	Observed (competitor evidence)	Latency is a Journey-1 failure mode
First-party tools shipped Q2 2026 (Claude Code Channels, Copilot Remote Control, Codex Mobile)	research/competitive-analysis.md	Dated competitor shipments; churn-threat inference	High	Observed (competitor evidence)	Drives churn-defense gap (Gap 4)
Must be under $30/mo; DIY, T3 Code, Emdash are free	research/icp.md + competitive-analysis	Price-sensitivity quote drives pricing band	High	Observed (user + competitor evidence)	Bounds Conversion / First-Bill pricing
Two tracks: Track A BYO-client + Track B Agent Board converge on managed infra	Concept briefs (personal + SaaS)	Product-design choice, not observed user behavior	Medium	Inferred (product framing)	Structures every journey and onboarding path
Discovery via blog / Show HN / Reddit / Discord peer rec	Indirect (ICP channel norms)	Channel inference for indie-hacker persona	Medium	Inferred (no direct funnel data)	Sets Discovery metrics (traffic source, CTR)
Free → paid conversion at free-tier limits; $9-29/mo flat	Concept brief (SaaS) pricing hypothesis	Conversion trigger inferred, not measured	Low	Inferred (no conversion data yet)	Conversion stage + First-Bill moment depend on it
Transaction: flat sub, no credit pricing, devbox add-on $49-199/mo	Concept brief (SaaS); Intent anti-pattern	Pricing-model design choice	Low	Inferred (provisional pricing)	Transaction stage economics (MRR, ARPU)
Retention: Agent Board becomes daily primary interface; switching cost	Inferred from persistence + mobile value	Retention loop assumed from aha value	Low	Inferred (no usage data)	Retention metrics (DAU/MAU, sessions/day)
Expansion: solo → team tier ($29-79/seat), land-and-expand	Concept brief (SaaS) growth hypothesis	PLG motion assumed for solo persona	Low	Inferred (no expansion data)	Expansion stage (NRR, seat expansion)
Advocacy via blog / HN / GitHub star / Discord; Recovery via re-engagement	Inferred from indie-hacker community norms	Advocacy + recovery loops assumed	Low	Inferred (no advocacy/win-back data)	Advocacy + Recovery metrics (NPS, referral, win-back)

Research Completeness

Is the evidence base complete enough to treat this journey map as approval-ready research, or is a specific evidence gap blocking?

Complete — observed evidence (aha, persistence, review bottleneck, pricing) is enough to proceed. Blocked — need real conversion / pricing data before the Conversion + Transaction rows can be trusted. Blocked — need real retention / expansion usage data before those stages are trusted. Blocked — need real discovery/funnel data (traffic source, CTR) before Discovery is trusted. Other / None of the above Need clarification

Observed vs Inferred Stages

Several later-lifecycle stages (Conversion, Transaction, Retention, Expansion, Advocacy, Recovery) and the two-track framing are inferred without direct user evidence. Are the inferred stages acceptable as-is, or do specific stages need real evidence before they ship into research/journey-map.md?

Acceptable — ship inferred stages clearly labeled as provisional; revise as data arrives. Conversion + Transaction must be evidence-backed before shipping (pricing is too load-bearing). Retention + Expansion + Advocacy + Recovery must be evidence-backed before shipping. The two-track framing itself must be validated with users before shipping. Other / None of the above Need clarification

Confidence & Assumption Register

A confidence read for every journey stage, flagging evidence-backed vs provisional/inferred, including each stage inferred rather than observed.

Item	Status	Confidence	What would change it
Trigger (laptop maxes out at 3-5 agents; mobile access need)	Evidence-backed	High	ICP interviews contradicting the scaling pain
Discovery (blog / HN / Reddit / Discord)	Provisional/inferred	Medium	Real funnel data showing a different dominant channel
Evaluation (free signup, compare to DIY, will not pay before trying)	Evidence-backed	High	Users reporting willingness to pay pre-trial
Onboarding (<5 min to first agent, both tracks)	Evidence-backed (bar) / inferred (track split)	Medium	Real onboarding time-to-value measurements
Aha Moment (laptop close + mobile check)	Evidence-backed	High	Users not recognizing this as the aha in testing
Conversion (free → paid at tier limits, $9-29/mo)	Provisional/inferred	Low	Measured conversion rate and price-point A/B data
Transaction (flat sub, no credits, devbox add-on $49-199/mo)	Provisional/inferred	Low	Billing experiments; willingness-to-pay for add-on
Retention (daily Agent Board, switching cost)	Provisional/inferred	Low	Real DAU/MAU and sessions/day from live usage
Expansion (solo → team tier $29-79/seat)	Provisional/inferred	Low	Observed team-tier upgrades and seat growth
Advocacy (blog / HN / GitHub star / Discord)	Provisional/inferred	Low	Measured referral rate and NPS
Churn (first-party tools close the gap; workflow change)	Provisional/inferred (threat is observed)	Medium	Exit-survey data on actual churn reasons
Recovery (feature re-engagement, lifecycle email, community)	Provisional/inferred	Low	Measured win-back rate from re-engagement campaigns
Two-track model (BYO-client + Agent Board)	Provisional (product framing)	Medium	User testing showing the track split confuses or splits demand

Provisional Assumptions

The low-confidence, inferred stages (Conversion, Transaction, Retention, Expansion, Advocacy, Recovery) carry the most assumption risk. How should they be treated in the shipped journey map?

Ship them labeled provisional/inferred, with the confidence column intact. Defer detail to the stage skills (/conversion-map, /retention-map, /expansion-map) and keep this map high-level. Drop the low-confidence stages from the map until evidence exists. Other / None of the above Need clarification

Scope & Non-Goals

Which tracks and personas this journey map covers, and which it deliberately leaves out.

In scope: The Agent Conductor persona (solo developer / indie hacker running 2-10+ AI coding agents), across Track A (BYO-client: T3 Code, Multica, Claude Code CLI connecting to managed infra, with Agent Board as infrastructure dashboard) and Track B (Agent Board as primary UI for users without an existing tool preference). Both tracks converge on gblock-party-managed server infrastructure.

Out of scope (for now): Team-lead / multi-seat buyer persona (only appears at the Expansion stage), enterprise procurement, and any self-hosted / BYO-server path. Stage-level detail (onboarding, conversion, transaction, retention, expansion) is delegated to the dedicated stage skills, not this overview.

Persona & Track Scope

Is the in-scope / out-of-scope split right for this map?

Yes — solo Agent Conductor across both tracks in scope; team/enterprise out for now. Add the team-lead persona to scope now, not just at Expansion. Narrow to Track A (BYO-client) only for the first map; defer Track B. Narrow to Track B (Agent Board) only for the first map; defer Track A. Other / None of the above Need clarification

Proposed File Changes

On approval, this map is written to the canonical journey-map research artifacts.

research/journey-map.md — canonical lifecycle overview: two-track user journeys, the full Trigger→Recovery lifecycle table, five critical moments, four journey gaps, evidence matrix, and confidence register. References the stage skills for deeper detail.
research/journey-map-interview.md — decision log capturing the compiled gate answers and section feedback from this alignment page.

Write Canonical Artifact

Approve writing this journey map to research/journey-map.md (and the decision log to research/journey-map-interview.md)?

Approve — write both research/journey-map.md and research/journey-map-interview.md. Write research/journey-map.md only; skip the interview/decision log. Hold — do not write canonical files until the flagged inferred stages have evidence. Other / None of the above Need clarification

Next Steps

Pick one:

/positioning (Recommended) — Positioning needs ICP, competitive analysis, and journey evidence, all of which now exist. The business-discovery pack is already enabled. Define competitive alternatives, unique attributes, value, target segment, and market category.
/onboarding-map — If the "First 5 Minutes" critical moment feels like the highest-risk stage to get wrong, map the onboarding flow in detail before positioning.

Post-Approval Route

Once this journey map is approved, which step comes next?

/positioning — ICP, competitive analysis, and journey evidence are ready. /onboarding-map — the "First 5 Minutes" moment is the highest-risk stage to get right first. /conversion-map — resolve the low-confidence Conversion / First-Bill economics first. Other / None of the above Need clarification

Compile Review

Use Compile Feedback to send concerns or clarification requests before answering every gate, or Compile Answers for final approval once all required gate questions are answered. The compiled YAML always includes any optional section feedback you set.