Archive/State of the Index — July 2026

Monthly Report

State of the Index — July 2026

Published August 1, 2026·RSS

The Ghost Index closed July at 55.7% — down 2.8 percentage points from 58.5% on July 2, on a base that grew by 16 net new agents across the month. The pattern is consistent with prior months: the index gets more agents before it gets more evidence. New entrants arrive without corroborated activity signals; they sit as ghosts on the record until they produce evidence or the window expires. That is the number working correctly, not a system under stress. The Developer category drove the headline — 717 alive of 1,319 indexed (54.4%), down from 56.6% at the end of W30. Service held at 83.3% (45 of 54). MCP servers and model families both stayed at 100%, both with the selection notes we carry honestly on every report.

The evidence-ranked tier expanded from roughly 145 agents at the start of July to 162 at month close — 17 promotions over the month. The most visible: CrewAI crossed to the top of the Developer ranking after satisfying the evidence gate on observable signals (47k GitHub stars, four confirmed ecosystem relationships, a verified on-chain identity). Tier promotion is not editorial discretion; it is a threshold crossing on published methodology. Seventeen agents crossed it in July.

On the other boards: the Model Family ranking was perfectly stable through the month — same five families, same scores, same order. The Service board was equally static. The Tokenized board saw minor score fluctuations of ±1 point, consistent with market moves rather than methodology changes. The archive compounded to 91,114 daily snapshots as of August 1, Merkle-anchored on Base every night. Each snapshot is the historical record nobody can backfill; the archive is the product.

All four boards at July close

Live from /api/rankings/*/llm-summary as of August 1, 2026.

Developer

GitHub · package usage · ecosystem signal
v2.c-public · full ranking →
1CrewAI75
2openclaw73
3openai-agents-python72
4DSPy Agents68
5google-adk-python67

The month's most consequential move: CrewAI crossed to #1 after being confirmed on observable signals — 47k GitHub stars, four ecosystem relationships, and a verified on-chain identity satisfied the evidence gate in mid-July. OpenClaw fell from 76 to 73 and dropped to #2. The rest of the board held through the month.

Model Families

HuggingFace · LMArena · deployment breadth
v1.4 · full ranking →
1Alibaba Qwen82
2Google Gemini80
3Mistral74
4DeepSeek74
5Meta Llama70

The most stable board we publish — same five, same order, same scores as the start of July. Qwen at 82, Gemini at 80, Mistral and DeepSeek tied at 74, Llama at 70. Open-weight families hold four of five seats and the argument stays unchanged: deployment breadth compounds where brand does not.

Tokenized

market cap · liquidity · holder basket
v1.1-tvl · full ranking →
1AIXBT81
2Ribbita73
3G.A.M.E65
4Luna64
5Vader60

AIXBT holds #1 at 81 (down 1 from July start). Ribbita moved up 1 to 73. G.A.M.E, G.A.M.E and Vader each down 1. Luna held at 64. The liveness column still reads 0% — the instrument probes HTTP endpoints while these agents live on-chain. The Ghost Index entry below carries the full flag.

Service

adoption · source quality · activity
v1.1-forks · full ranking →
1A2A77
2a2a-python74
3evolver73
4a2a-samples72
5bitterbot-desktop70

No movement across July. A2A and three of its reference implementations hold the top four, packed inside five points; bitterbot-desktop at 70. The service board closed July exactly as it opened it.

Ghost Index: 55.7% — month closed 3 points lower

55.7%

787 alive · 626 ghosts · 1,413 indexed · July 2 open: 58.5%

−2.8 points across July. The index grew by 16 agents; the alive share declined. The archive is nightly; every number here is independently recomputable via /oracle.

MCP servers*15 / 15100%
Model families10 / 10100%
Service45 / 5483.3%
Developer717 / 131954.4%
Tokenized*0 / 150%

* 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.

July 2026 in data

1,413

Agents indexed

55.7%

Ghost Index liveness

91,114

Snapshots archived

162

Evidence-ranked (July 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.

All Rankings →Ghost Index →Daily Changes →Methodology →RSS →