I. Thesis
A directory and community platform for people building AI agents — vibecoder developers, indie hackers, and enterprise builders shipping agentic software.1 Purpose: tap into the agentic ecosystem for distribution and community, connecting Eric to builders who could become users of his other products (Donna, avet, Blackring).1
Originally conceived as a “Vibecoder Directory” in collaboration with David Li (VR) — scraping Lovable projects and researching existing directories. Pivoted Feb 6 to broader “Agent Creator Directory.”1
The honest framing: This isn’t a product solving a user pain point. It’s a distribution play — Eric wants to be visible in the agentic ecosystem. The question is whether building a directory is the right vehicle for that.
II. Dog-Food Signal
Weak Dog-Food — Eric Doesn’t Use Directories
- How does Eric actually find agent builders? Through coffee chats, WhatsApp groups, referrals, and natural network effects.1
- Has Eric ever opened an agent directory to discover someone? No evidence in CRM data. He met Conrad Ho through school, Wenhao through friends, David Li through GenieFriends, Leo Ho through Ronald Yao’s intro.1
- Does Eric have the pain of “I can’t find agent builders”? No. He has 50+ active contacts, many of whom are building or interested in agents.1
- Contrast with Agent Elo: Eric genuinely asks “Which agent should I use for X?” — that’s dog-food. “Where do I find agent builders?” — he already knows where. No acute pain.
III. Market Sizing
| Layer |
Market Size |
Source |
| Global AI Agent Market (2025) |
US$7.63B |
MMNTM2 |
| Global AI Agent Market (2033) |
US$183B (CAGR 49.6%) |
MMNTM2 |
| AI Tool Directory Market |
~US$100–500M (top 5 directories reach 15M MAU)3 |
MKT Clarity, Techraisal |
| Developer Community Platforms |
US$1–3B (GitHub: $2.8B, GitLab: $860M ARR)4 |
Fueler, GitLab SEC |
| Agent-Specific Directories (est.) |
US$10–50M |
Estimated from existing players |
| Eric’s Addressable (Year 1) |
US$0–6K/year |
Calculated below |
Why “Agent Directory” Is Not a Recognised Market Segment
No research firm tracks “AI agent creator directories” as a category. The TAM borrows from adjacent markets: developer tools ($2.8B GitHub), AI tool discovery ($100–500M), and developer communities ($1–3B). The actual addressable slice for a solo founder building a directory from scratch is microscopic.3
IV. The Landscape — It’s Already Crowded
This is the core finding. The space is saturated. There are 100+ AI tool directories, and agent-specific directories are proliferating fast.3
Agent & AI Directories (Direct Competitors)
| Platform |
Traffic/Size |
Revenue |
Model |
| AI Agents Directory5 |
100K MAU, 23K newsletter |
$29–500/listing + newsletter |
Listings + ads + newsletter |
| Agent.ai6 |
343K visits/mo, 1,600+ agents |
Undisclosed |
Marketplace transactions |
| Cursor.directory7 |
71K members, 250K MAU |
$35K/month |
Featured listings + MCP directory |
| Toolify.ai8 |
26K+ tools, 450+ categories |
Ads + submissions |
Display ads + paid submissions |
| Vibecoding.builders9 |
62+ builders, 97+ projects |
Free (community) |
Free showcase, no monetization |
| AI Builders Community10 |
Undisclosed |
$10/mo membership |
Paid community (Skool) |
| AI Agent Store11 |
750+ agents |
Undisclosed |
Curated directory + metrics |
MCP Server Directories (Infrastructure Layer)
The 2000s Web Directory Déjà Vu17
- A widely-shared Dev.to article draws direct parallels between the AI directory boom and the 2000s web directory crash: “same mechanics, same inevitable outcome.”
- No meaningful traffic: Most AI directories attract visitors but fail to convert them into actual customers17
- Diluted SEO value: With millions of outgoing links, individual backlinks provide negligible benefit17
- User behavior: Customers prefer Google, AppSumo, or trusted reviews over directories to find tools17
- Directory fatigue is real — the question isn’t “does the world need another directory?” but “why would this one win?”
Why Cursor.directory Succeeded (and Why Eric Can’t Replicate It)
Cursor.directory is the benchmark: solo founder, $35K/month, bootstrapped.7 But it succeeded because of three structural advantages Eric doesn’t have:
- Tool-specific gravity. It’s the community hub for ONE tool (Cursor) that millions of developers use. Eric’s directory has no single tool to orbit.
- MCP timing. Cursor.directory became the de facto MCP server browser right as MCP adoption exploded. That window has passed — the official MCP Registry is now backed by Anthropic, GitHub, and Microsoft.12
- SEO head start. Launched 2024, built 1+ years of domain authority and backlinks. Starting from zero in Feb 2026 means racing against entrenched competitors.
V. Failed Examples — The Graveyard
| Example |
What They Tried |
What Killed It |
Lesson |
| GPT Store18 |
3M custom GPTs, OpenAI-backed marketplace |
No quality signal, terrible discovery, agents disappear from search |
Even with OpenAI’s scale, directory-without-ranking fails |
| ChatGPT Plugins |
Tool-use marketplace for ChatGPT |
Shut down entirely — no quality signal = chaos |
Tool discovery without ranking doesn’t work |
| Smartlinks.ai19 |
Centralised backlink exchange platform |
Shut down March 2024. 2,000+ website requests but no sustainable traction |
Directory ≠ product-market fit |
| Dev.to Listings20 |
Classified listings within dev community |
Discontinued — couldn’t invest enough to compete |
Even established communities struggle to monetise listings |
| Paid Community Pattern21 |
$39/mo community membership (B2C) |
$4,650 total revenue over 2 years. Founder quit. |
B2C community monetisation is brutally hard |
| Fixie.ai |
Agent marketplace → pivoted to enterprise |
Consumer agent marketplace had zero adoption |
B2C agent discovery didn’t work |
Death Pattern: Directory Without Differentiation
Every failed directory shares the same root cause: no structural moat. Anyone can copy a directory. The data is public. The listings are commodity. The only sustainable directories are ones attached to something harder to replicate — a tool (Cursor), a community ritual (Product Hunt daily launches), or a quality signal (Elo ranking).17
VI. Unit Economics (Benchmarked)
Revenue Benchmarks for Solo-Founder Directories
| Platform |
Monthly Revenue |
How |
Time to Get There |
| Cursor.directory7 |
$35,000 |
Featured listings, MCP directory, 250K MAU |
~1–2 years + Cursor’s growth wave |
| SaaSHub22 |
$10,000+ |
106 featured listings @ $130/mo, 1.7M visitors |
2+ years of SEO investment |
| LibHunt22 |
$10,000+ |
Display ads, 1.7M monthly visitors |
3+ years of content |
| Product Hunt23 |
$325K (annualised) |
Ads + direct sales, 77-person team |
10+ years (acquired for $20M) |
| AI Agents Directory5 |
Est. $5–15K |
$29–500 listings + $500/wk banners + newsletter |
~1 year (Microsoft for Startups backing) |
Eric’s Projected Economics
Year 1 Rev
$0–6K
$0–500/mo
Year 2 Rev
$12–36K
$1–3K/mo (if consistent)
Time Required
10–20
hrs/week (content + SEO)
Opp. Cost
$192K+
vs. Sourcy + Agent Elo
COGS Breakdown
| Cost Component |
Monthly Cost |
Notes |
| Hosting (Vercel/Cloudflare) |
$0–20 |
Static site, minimal infra |
| Domain |
$1 |
~$12/year |
| Content creation |
$0 cash / 40–80 hrs |
THE REAL COST. Eric’s time. |
| SEO tools |
$0–100 |
Ahrefs/Semrush if serious about SEO |
The Death Metric: Opportunity Cost of Eric’s Time
Eric bills HK$16K cash + HK$32K equity/month at Sourcy.1 That’s ~HK$48K/month for ~40 hrs/week of work. Every 10 hours spent on a directory that makes $0–500/month is HK$12,000 in lost Sourcy time.
Worse: those same 10 hours could go to Agent Elo (which has a clear path to $1–10K/month) or Donna pilot expansion (which builds enterprise value).
A directory must earn >$1,500/month to justify 10 hrs/week of Eric’s time. Based on benchmarks, that takes 12–24 months of consistent effort. Negative ROI for at least a year.
VII. Live Signals (February 2026)
Directory Fatigue Is Setting In3
- A 2025 Directory Trends Report analysed 50+ directories to identify “real revenue opportunities” — the fact this report exists tells you the space is saturated24
- RankMyAI expanded to 46,000 AI tools in its database25
- There’s An AI For That lists 12,800+ tools and charges $347 for a permanent listing3
- Top 5 AI directories collectively reach 15 million monthly visitors3
But Agent Builders ARE Congregating
- Cursor.directory: 71.4K members, 250K MAU, $35K/mo revenue — proving developer community around agentic tools works7
- Vibecoding.builders: 62+ builders actively showcasing projects built with AI tools9
- MCP ecosystem explosion: 5,000+ servers, 6.6M monthly SDK downloads, official registry now backed by Anthropic/GitHub/Microsoft12
- HN interest: “Show HN: I manually curated 130 modern AI Agents” received engagement11
The signal is clear: demand for connecting with agent builders exists. But that demand is being met by existing platforms, not by a gap waiting for a new directory. The question is whether Eric brings something they don’t.
VIII. GTM Assessment — Founder-Contextualised
Eric’s Advantages for a Directory
- Builder credibility: Shipping Donna, avet, OpenClaw infrastructure, Blackring. Not a curator — a practitioner.1
- Network access: 50+ active contacts, many technically oriented or building agents.1
- Content potential: Deep technical context from running multiple agentic projects simultaneously.
Eric’s Disadvantages for a Directory
- No SEO moat: Starting from zero domain authority against competitors with 1–3 years of content. Structural disadvantage.
- No content creation habit: Eric’s strengths are building products and networking, not writing weekly blog posts or maintaining curated lists.1
- Bandwidth: Sourcy (high priority), Blackring (high priority), Donna pilots, Agent Elo. Build deficit day 5.1
- No viral hook: Cursor.directory rode Cursor’s growth. What growth wave would this directory ride?
- Johnny cooling off: Subvert Johnny “not engaged, not responding recently” — complaining projects are “so far away from revenue.”1 Collaborator attrition risk.
The Merge Question: Directory ∈ Agent Elo?
YES — The Directory Is a Feature, Not a Product
- Agent Elo already needs a “builder profiles” layer. When builders submit agents for ranking, they create profiles. That IS the directory — emergent, not built separately.
- The directory brings traffic to Agent Elo. SEO for “who’s building AI agents” feeds into “which agents are best?” Funnel, not standalone.
- Distribution for Eric’s projects (the original goal) works better through Agent Elo. Donna and avet as ranked agents = credibility. A separate directory listing them = just another link.
- The value Eric wants — connecting with agent builders — happens as a byproduct of running Agent Elo, not as a purpose-built directory.
Government Grants
Not applicable. A directory/community platform doesn’t qualify for PSG, EDG, or IMDA grants. No R&D component, no technology innovation, no enterprise transformation angle. No grant lever.
IX. Red Team
Why Build the Directory
- Agent builders are a growing audience (MCP: 5K+ servers, 6.6M SDK downloads/mo)12
- Cursor.directory proves solo founder can hit $35K/mo7
- Eric has builder credibility (Donna, avet, OpenClaw)
- Low COGS ($20/mo hosting) = high gross margin if it works
- Content flywheel: builders submit, content grows, SEO improves
- Network effects: more builders → more visitors → more builders
Why Skip the Directory
- 100+ AI directories already exist — severe saturation3
- 2000s web directory crash pattern repeating17
- Official MCP Registry backed by Anthropic/GitHub/Microsoft12
- Eric doesn’t use directories himself (no dog-food)1
- Revenue ceiling for solo founder: $10–35K/mo after 1–2 years7
- 12–24 months to break even on time investment
- Content/SEO is not Eric’s strength or habit
- Bandwidth already maxed (Sourcy, Blackring, Donna, Elo)
- No viral hook to ride (unlike Cursor.directory → Cursor)
- Directory has zero switching costs — no lock-in
- Johnny (collaborator) cooling off
- The distribution goal is better served by Agent Elo
Steel-Man: The Best Version of This Idea
“Forget generic directory. Build an opinionated, curated weekly newsletter of the 5 best agent projects — Eric’s picks. Short-form, high-taste, zero SEO dependency.”
This is the strongest version of the directory concept because:
- Curation is Eric’s actual skill — he evaluates opportunities (see: this research system). Taste-driven curation > exhaustive listing.
- Newsletter ≠ SEO business. Doesn’t require 12–24 months of domain authority building. Grows through referrals + social.
- 2 hours/week, not 20. Sustainable alongside other priorities.
- Feeds Agent Elo directly. “Best agents this week” → natural segue to “ranked on Agent Elo.”
But even this is better positioned as the Agent Elo newsletter, not a standalone directory brand. Feature, not product.
Verdict: SKIP as Standalone — Absorb into Agent Elo
A standalone agent creator directory is a bad use of Eric’s time.
The evidence is overwhelming:
- 100+ AI directories already exist. The space is saturated and showing 2000s web directory crash patterns.17
- No dog-food signal. Eric doesn’t use directories to find builders. He uses his network.
- Revenue ceiling is low: $10–35K/month after 1–2 years of consistent effort, vs. Agent Elo which could reach the same range faster with a clearer value prop.
- Official MCP Registry is now backed by Anthropic, GitHub, and Microsoft.12 You can’t out-directory the protocol authors.
- Opportunity cost is devastating: Every 10 hours on a directory = HK$12K in lost Sourcy time + delayed Agent Elo + delayed Donna pilots.
- The distribution goal — connecting with agent builders — is better served by Agent Elo (builders submit agents → natural community) or by content (HN posts, LinkedIn) with zero infrastructure needed.
What to do instead:
- Build Agent Elo. The “builder profiles” layer emerges naturally when people submit agents. That IS the directory.
- Post to HN/LinkedIn. Eric’s builder credibility (Donna, avet, OpenClaw work) generates more ecosystem connections than any directory ever would.
- Optional: “Eric’s Agent Picks” newsletter. If Eric wants a curation outlet, 2 hrs/week curating 5 best agents and shipping a newsletter is 10x more capital-efficient than building a directory.
- Let VR (David Li) run it if he wants. The collaboration was VR-initiated. If VR wants to build a directory, Eric can contribute his network and appear as an advisor without committing build time.
The one thing that would change this verdict: If Eric discovers a structural differentiation that no existing directory has — something like “live agent performance data” or “trust scores from community usage.” But that’s Agent Elo, not a directory. The convergence is the answer.
References
[1] Eric’s CRM state files — projects.json, people.json, daily reports (Jan–Feb 2026). Agent Creator Directory project notes, contact context, bandwidth assessment
[2]
MMNTM — The Top 100 AI Agent Companies: A Strategic Directory AI agent market: $7.63B (2025), $183B (2033), 49.6% CAGR. 57% of companies have agents in production.[3]
MKT Clarity — List of Directories to Submit Your AI Tool 15M+ monthly visitors across top 5 AI directories. TAAFT: 12,800+ tools, $347/listing. Toolify: 26K+ tools, $49–99/listing.[4]
Fueler — GitHub in 2026: Usage, Revenue, Valuation GitHub projected $2.8B revenue 2026, 100M+ developers, 90% from enterprise.[5]
AI Agents Directory — About 100K+ monthly users, 23K newsletter subscribers, $29–500 pricing tiers, Microsoft for Startups backing.[6]
Semrush — agent.ai traffic stats 343K visits/mo (Dec 2025), 1,600+ agents, global rank 117,763.[7]
Tiny Startups — cursor.directory revenue $35,000/month, solo founder, bootstrapped, Sweden-based. 71.4K members, 250K MAU.[8]
Techraisal — Toolify.ai breakdown 26,000+ tools, 450+ categories. Ads + developer submissions. Strong SEO visibility.[9]
Vibecoding.builders 62+ active builders, 97+ projects, free showcase platform. No monetisation.[10]
The AI Builders Community $10/mo membership, StarterKit included, live vibe coding sessions.[11]
HN — Show HN: I manually curated 130 modern AI Agents Community interest in curated agent directories (Aug 2024).[12]
The MCP Registry (official) Canonical registry backed by Anthropic, GitHub, PulseMCP, Microsoft. REST API, namespace management.[13]
MCPMarket 631+ pages of MCP servers, community-maintained marketplace.[14]
MCP Server Directory 2,500+ MCP resources across multiple categories.[15]
Open MCP Directory (OMDR) Trust-verified MCP servers. Security verification, 90% creator revenue share. Free local install + hosted options.[16]
LobeHub MCP Marketplace Community-rated MCP servers. Activity, stability, and community feedback scoring.[17]
Dev.to — The AI Directory Craze is a 2000s Web Directory Déjà Vu Directory fatigue analysis. No meaningful traffic conversion. Diluted SEO value. Users prefer Google.[18]
OpenAI Community — GPT Store discovery problems GPT Store failure: 3M GPTs, terrible discovery, agents disappearing from search.[19]
Medium — Shutting down Smartlinks.ai Directory/backlink exchange platform. 2,000+ requests but no sustainable traction. Shut down March 2024.[20]
Dev.to — We Are Turning off Listings Dev.to discontinued Listings feature. Redirected to Billboards system.[21]
IndieHackers — Don’t make a paid community $4,650 total revenue over 2 years from $39/mo community. B2C community monetisation failure case.[22]
eBizFacts — SaaSHub & LibHunt earning $10K+/month SaaSHub: $10K+/mo from 106 featured listings @ $130/mo. LibHunt: $10K+/mo from display ads, 1.7M visitors.[23]
Latka — Product Hunt revenue $3.9M revenue (2024), 76% YoY growth, 77-person team. Acquired for $20M (2016).[24]
DirectoryIdeas — Directory Trends Report 2025 50+ directories analysed for revenue opportunities. Existence of this report signals market saturation awareness.[25]
RankMyAI — Expanding to 46,000 AI tools World’s largest AI tools directory. Scale makes new entrants extremely difficult to differentiate.