Archive/State of the Index — August 2026
Monthly Report
State of the Index — August 2026
The Ghost Index opened August at 55.7%, briefly peaked at 56.4% in the first week, then declined steadily through the month to close at 54.0% on August 31 — a net loss of 1.7 percentage points. The indexed base grew by 21 agents (1,413 → 1,434). The pattern holds as it has every month: new agents enter the index before they produce corroborated activity signals; they sit as ghosts on the record while the window runs. That is the system working correctly. The Developer category drives the headline as before — 691 alive of 1,326 indexed (52.1%), with Service holding well at 86.8% (59 of 68).
The evidence-ranked tier expanded from 162 at July close to 191 at August close — 29 promotions over the month. No single promotion carried the narrative weight of July's CrewAI crossing, but the pace held steady. On the Developer board, the month's notable move was openclaw falling from #2 to #3 as OpenAI Agents Python climbed from #3 to #2. The Model Family board compressed slightly — same five families, same order, scores down 1–2 points each — consistent with LMArena ranking fluctuations rather than methodology changes. The Service board saw one new entry: agent-teams-ai appeared at #5 as a2a-samples dropped out of the top five.
The archive compounded to 135,267 daily snapshots as of September 1 — 44,153 new records added in August alone, every one Merkle-anchored to Base at 04:30 UTC the following morning. August adds to an unbroken run of nightly snapshots since April 2026. Each snapshot is a timestamped record of the agent economy that cannot be backfilled; the longitudinal archive is the product and it compounded again this month without interruption.
All four boards at August close
Live from /api/rankings/*/llm-summary as of September 1, 2026.
Developer
GitHub · package usage · ecosystem signalThe month's notable position move: openclaw fell from #2 to #3 as OpenAI Agents Python climbed from #3 to #2. CrewAI holds #1 at 74, down 1 point from July's 75 — the evidence gate keeps scores anchored to signals, not momentum. DSPy and Google ADK swapped positions (DSPy fell from #4 to #5), ending August level on score at 67.
Model Families
HuggingFace · LMArena · deployment breadthSame five families, same order. Scores compressed slightly: Gemini 80→79, Mistral 74→73, DeepSeek 74→72; Qwen and Llama held at 82 and 70. The board reflects Qwen's derivative count advantage (1,046 fine-tuned variants tracked) against Gemini's deployment breadth lead (145 tracked deployments). No methodology changes in August.
Tokenized
market cap · liquidity · TVLAIXBT holds #1 at 80 (−1 from July). G.A.M.E and Luna each gained 1 point (65→66, 64→65). Ribbita and Vader held. The liveness column still reads 0% — a standing instrument gap: the probe listens on HTTP while these agents live on-chain. The flag is noted below.
Service
adoption · source quality · activitya2a-samples dropped out of the top 5 (was #4 in July at 72). bitterbot-desktop rose from #5 to #4. agent-teams-ai entered at #5 with 69. The A2A protocol cluster still holds the top three; the board remains tightly packed with 8 points separating #1 from #5.
Ghost Index: 54.0% — closed 1.7 points below July
54.0%
775 alive · 659 ghosts · 1,434 indexed · Aug 1 open: 55.7%
−1.7 points across August. Brief peak at 56.4% (Aug 7), then a steady two-week decline. Every number here is independently recomputable via /oracle.
* Two standing flags. MCP's 100% reflects a selection effect — we track actively maintained servers. Tokenized's 0% is an instrument gap: the probe listens on HTTP while those agents live on-chain. The fix is on the roadmap; when the number moves it will move because the measurement improved, stated in those words.
August 2026 in data
1,434
Agents indexed
54.0%
Ghost Index liveness
135,267
Snapshots archived
191
Evidence-ranked (Aug close)
The index runs autonomously — daily snapshots at 02:00 UTC, nightly liveness scoring at 23:50 UTC, Sunday ranking runs, and Merkle anchoring to Base at 04:30 UTC. Every number here is live and independently recomputable via /oracle. The archive is the product: each nightly snapshot is a timestamped record of the agent economy that cannot be backfilled. The data is available machine-readable at /api/ghost-index/v1, /api/rankings/*/llm-summary, and /api/agent-economy/llm-summary.