Key Takeaways
- Only 2 new projects added in May (Darwin, ClawVault) vs 37 in April — a deliberate slowdown
- +164,546 stars across 173 tracked repos (+12.4% in 29 days) — the strongest growth month on record
- Hermes Agent broke 150K then 170K in the same month (+44,955 stars, +36%)
- Skill hubs are eating spotlight: Imbad0202/academic-research-skills jumped +491% (3,914 → 23,130), the biggest single move of the month
- Academic writing sub-category heating up: Imbad0202, Master-cai (+71%), blader/humanizer (+29%), Leey21 (+26%) all surging
- Ecosystem total crossed 1.5 million stars on 2026-05-27
- 18 projects now marked
stable(up from 17 in April) — paper-locked / release-driven projects no longer trigger dormant alerts
The Numbers
| Metric | End of April | End of May (5/28) | Δ |
|---|---|---|---|
| Tracked projects | 144 | 146 | +2 |
| Skill hubs | 33 | 33 | 0 |
Projects marked stable | 17 | 18 | +1 |
Projects marked misleading | 0 | 1 | +1 |
| Projects in graveyard | 1 | 1 | 0 |
| Total stars (173 tracked, 30-day window) | 1,328,484 | 1,493,030 | +164,546 (+12.4%) |
This is the first month with a true ~30-day stars window (4/30 → 5/28), as promised in last month's methodology note.
Top Movers — Projects (4/30 → 5/28)
By absolute stars added:
| Project | 4/30 | 5/28 | Added | % |
|---|---|---|---|---|
| NousResearch/hermes-agent | 126,019 | 170,974 | +44,955 | +36% |
| anthropics/skills | 126,421 | 142,326 | +15,905 | +13% |
| multica-ai/multica | 23,096 | 33,865 | +10,769 | +47% |
| openclaw/openclaw | 366,623 | 375,222 | +8,599 | +2% |
| karpathy/autoresearch | 78,015 | 83,840 | +5,825 | +7% |
By percentage (start ≥ 50 stars):
| Project | 4/30 | 5/28 | Added | % |
|---|---|---|---|---|
| Imbad0202/academic-research-skills | 3,914 | 23,130 | +19,216 | +491% |
| openags/OpenAGS | 60 | 284 | +224 | +373% |
| reacher-z/ClawBench | 120 | 342 | +222 | +185% |
| Yuan-lab-LLM/ClawManager | 620 | 1,443 | +823 | +133% |
| aipoch/medical-research-skills | 398 | 877 | +479 | +120% |
Imbad0202 quintupling in a month is the May headline. Combined with Master-cai (+71%) and blader/humanizer (+29%), the "academic writing skills" sub-category is accelerating faster than autonomous research itself.
The May Story: Consolidation, Not Expansion
April added 37 new projects in 30 days. May added 2. That isn't a slowdown in the field — it's a slowdown in us. Three reasons:
- The April rush was front-loaded by the autoresearch wave (6 projects in one curation pass on 4/14). May didn't have an equivalent surge.
- Curation got more selective. We rejected
obclaw(6⭐, Obsidian notes) early in May for being non-scientific, and have raised the bar for what counts as "in scope." - Existing projects are absorbing attention. Hermes alone gained more stars in May (+45K) than the entire ecosystem net growth in March.
The implication: the OpenClaw ecosystem is in a consolidation phase. The top projects are pulling away from the long tail.
The Hermes Acceleration
Hermes Agent's May was extraordinary:
| Date | Stars | Δ |
|---|---|---|
| 4/30 | 126,019 | — |
| 5/11 | 143,914 | +17,895 |
| 5/17 | 154,009 | +10,095 (crossed 150K) |
| 5/22 | 162,684 | +8,675 |
| 5/27 | 169,571 | +6,887 |
| 5/28 | 170,974 | +1,403 (crossed 170K) |
That's ~20% of the entire ecosystem's stars in a single project, and most of it accumulated this month. The "agent harness" category — which we defined in our 2026-05-17 Hermes review — is becoming legible to the broader Claude / OpenClaw community, and Hermes is its de facto reference implementation.
We expect Hermes to cross 200K in June if current pace holds. (We'd happily be wrong about the deceleration we keep predicting.)
The Skill-Hub Explosion
Five skill hubs grew faster than the ecosystem average (+12%) in May:
- Imbad0202/academic-research-skills: +491% (3.9K → 23K)
- K-Dense-AI/claude-scientific-skills: +33% (19.8K → 26.3K)
- VoltAgent/awesome-agent-skills: +20% (19.5K → 23.4K)
- Leey21/awesome-ai-research-writing: +26% (20.5K → 25.9K)
- blader/humanizer: +29% (16.6K → 21.3K)
The shared theme: academic writing and research skills. We covered this in Biomedical AI Agents 2026 and the new Nature Skills deep dive, but the May numbers say the wave is much broader than one project — it's a category.
The Two New Arrivals
Darwin (yejunbin, added 2026-05-09): PRISMA-compliant biomedical research agent forked from feynman. 1⭐ at addition, methodologically serious (MIQE / CONSORT / ARRIVE compliance, HGNC / RRID / UniProt nomenclature). Reviewed in our Biomedical AI Agents 2026 comparison.
ClawVault (tophant-ai, added 2026-05-26): OpenClaw security vault. 1,220⭐ at addition (created March), sits in the security group alongside ClawKeeper and ClawSafety. Real product for science labs that need PHI / IRB / HIPAA compliance when running AI agents.
May's two arrivals are both about quality of methodology: Darwin enforces reporting standards, ClawVault enforces operational standards. The shift from "more projects" to "more rigorous projects" looks intentional rather than accidental.
Watching List
Two new entries to watch in June:
- qinheming/bioclaw — saturated at 103-105⭐ for most of May, then woke up 5/26: +24 stars in 4 days (105 → 129). Either the project is shipping again or someone influential just shared it. Worth a closer look.
- cuhk-aim-group/neuroclaw — went 14 → 53⭐ in May, neuroscience-focused. CUHK AIM Group brand carries weight; if it crosses 100⭐ in June we promote it from watching to featured.
What Went Quiet
anthropics/skills only grew 13% in May (vs Hermes 36%) — first month the official Anthropic skills repo grew slower than the broader ecosystem. Could be saturation (it's already 142K). Could be that community skill hubs are now where the marginal star is going.
What to Watch in June
- Hermes crossing 200K — if pace holds, mid-to-late June
- academic-research-skills crossing 50K — currently 23K and accelerating
- Will new-project additions stay under 5 per month? — does consolidation continue?
- Will any current "watching" repo cross our promote threshold? — bioclaw, neuroclaw are the candidates
- First AI Overview citation of a Claw4Science post — three GEO-heavy posts now live (Nature Skills, Biomedical Agents, Drug Discovery refresh); we'll know by July if the structural changes paid off
Methodology
- Star counts: GitHub stars via our
/api/cron-refreshendpoint, snapshotted daily intodocs/stars-history.csv - Project count: live from
src/config/locale/messages/en/pages/index.json(projects) andskill-hubs.json(hubs); growth comparisons separate the two - Growth window: 2026-04-30 → 2026-05-28 (29 days). First near-full-month comparison since data collection stabilised
- Top-mover filters: percentage rankings require start ≥ 50 stars to filter noise
stablestatus: human-curated for paper-anchored projects with verified Nature / arXiv / journal publications (BixBench, ChemCrow, Virtual Lab, MDCrow, HeurekaBench, OmniScientist, etc.)
Previous report: Ecosystem Report: April 2026 — Benchmarks and Autoresearch — 144 projects, +37 new, Hermes +44K, autoresearch cluster.
Last updated: 2026-05-28.
