State of Onchain Agents 2026

1. 01

   Verified installs

   39.9%

   of listings with install steps, reproduced in a container

   As of2026-10-19Sample313

   MethodShowHide

   Share of listings with a deploy spec whose install we reproduced in a container

   Stage

   directory / deploy_status=verified

   Method

   weekly container run of the published install steps

   Sample

   313

   As of

   2026-10-19

   Note

   125 of 313. Says the install steps complete; nothing about runtime or safety.

   Reproduce

   jq '[.[]|select(.deploy_spec!=null)]|length, ([.[]|select(.deploy_spec.deploy_status=="verified" or any(.deploy_spec.install_variants[]?; .deploy_status=="verified"))]|length)' data/resources.json

   Open this number→
2. 02

   Install test runs

   184 / 335

   install runs completed, out of all recorded

   As of2026-10-19Sample335

   MethodShowHide

   Container verification attempts recorded, and how many passed

   Stage

   deploy_verification_runs

   Method

   one row per (slug, ran_at, lane) from the weekly verifier

   Sample

   335

   As of

   2026-10-19

   Note

   A failing run names the rung it failed on; the queue retries failures after 28 days. A credentials rung means the spec needs a secret the harness does not hold, not that the install is broken.

   Reproduce

   select count(*) filter (where ok), count(*) from deploy_verification_runs;

   Open this number→
3. 03

   Daily-checked listings

   430

   listings with an observed record of our own daily checks

   As of2026-10-19Sample430

   MethodShowHide

   Listings with a 14-day observed record of our own checks

   Stage

   directory / uptime_observed.rate_days ≥ 14

   Method

   daily probe of the endpoint the listing publishes

   Sample

   430

   As of

   2026-10-19

   Note

   A success rate of OUR checks, not uptime: it covers only checks since 2026-09-01, when a bot-block stopped counting as a failure.

   Reproduce

   jq '[.[]|select(.uptime_observed.success_rate_pct!=null)]|length' data/resources.json

   Open this number→
4. 04

   Consistently answering

   88.6%

   of observed listings that answered nearly every check

   As of2026-10-19Sample430

   MethodShowHide

   Share of those answering ≥95% of our checks

   Stage

   directory / observed success ≥ 95

   Method

   same probe series

   Sample

   430

   As of

   2026-10-19

   Reproduce

   jq '[.[]|select(.uptime_observed.success_rate_pct>=95)]|length' data/resources.json

   Open this number→

Counts per chain do not add up. Each is its own stage.

1. 05

   Ethereum registrations

   54,306

   agent IDs registered. A registration is not a running agent

   As of2026-10-19

   MethodShowHide

   ERC-8004 agents registered on Ethereum

   Stage

   erc8004 / Ethereum / registered

   Method

   lastAgentId (_lastId, ERC-7201 namespaced storage slot) on the singleton Identity Registry (tier 1)

   Sample

   n/a

   As of

   2026-10-19

   Note

   Read from the registry's storage; a registration is a mint, not a running agent.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='erc8004' and chain='Ethereum' and stage='registered' and week='2026-W43';

   Open this number→
2. 06

   Base registrations

   271,519

   agent IDs registered. A registration is not a running agent

   As of2026-10-19

   MethodShowHide

   ERC-8004 agents registered on Base

   Stage

   erc8004 / Base / registered

   Method

   lastAgentId (_lastId, ERC-7201 namespaced storage slot) on the singleton Identity Registry (tier 1)

   Sample

   n/a

   As of

   2026-10-19

   Note

   Read from the registry's storage; a registration is a mint, not a running agent.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='erc8004' and chain='Base' and stage='registered' and week='2026-W43';

   Open this number→
3. 07

   BNB Chain registrations

   376,540

   agent IDs registered. A registration is not a running agent

   As of2026-10-19

   MethodShowHide

   ERC-8004 agents registered on BNB Chain

   Stage

   erc8004 / BNB Chain / registered

   Method

   lastAgentId (_lastId, ERC-7201 namespaced storage slot) on the singleton Identity Registry (tier 1)

   Sample

   n/a

   As of

   2026-10-19

   Note

   Read from the registry's storage; a registration is a mint, not a running agent.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='erc8004' and chain='BNB Chain' and stage='registered' and week='2026-W43';

   Open this number→
4. 08

   New Base registrations

   176,818

   new agent IDs registered since block 51,523,550

   As of2026-10-19

   MethodShowHide

   New ERC-8004 registrations on Base over the window

   Stage

   erc8004 / Base / registered_new_in_window

   Method

   lastAgentId at the head minus lastAgentId at block 51523550 — exact, whole population, no sampling (tier 1)

   Sample

   n/a

   As of

   2026-10-19

   Note

   From block-sampled storage deltas; the row states its window.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='erc8004' and chain='Base' and stage='registered_new_in_window' and week='2026-W43';

   Open this number→
5. 09

   Identity and payment overlap

   Unknown

   sampled smart accounts that hold an agent ID or paid an x402 seller

   As of2026-10-19

   MethodShowHide

   Sampled smart accounts that ALSO hold an ERC-8004 id or settle x402

   Stage

   erc4337_accounts / Base / intersected

   Method

   for up to 250 distinct accounts from the deployment sample: balanceOf(account) on the ERC-8004 Identity Registry, plus membership of the set of addresses that paid a catalogued x402 seller in the same window (tier 2)

   Sample

   0

   As of

   2026-10-19

   Note

   The intersection is the number, not either count alone. This is the only figure on this venue that may be described as adoption, and it is still an indicator rather than proof — an account can hold an agent id without being an agent. Both legs are one-week windows; an account that settled x402 last month does not appear.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='erc4337_accounts' and chain='Base' and stage='intersected' and week='2026-W43';

   Open this number→
6. 13

   Solana identity accounts

   1,696

   program accounts. At most this many agents

   As of2026-10-19

   MethodShowHide

   Solana agent-identity program accounts (UPPER BOUND)

   Stage

   solana_mpl_agent_identity / Solana / registered_upper_bound

   Method

   getProgramAccounts on 1DREGFgysWYxLnRnKQnwrxnJQeSMk2HmGaC6whw2B2p with a zero-length dataSlice (keys only) via public mainnet RPC; count of returned accounts (tier 2)

   Sample

   n/a

   As of

   2026-10-19

   Note

   getProgramAccounts counts every account the program owns, agents or not. An account is not an agent, and the count includes per-agent metadata PDAs and unused entries. Read as 'at most this many'. Solana x402 settlement is NOT covered by any row in this collector.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='solana_mpl_agent_identity' and chain='Solana' and stage='registered_upper_bound' and week='2026-W43';

   Open this number→
7. 15

   Top registering platform

   120

   sampled registrations from api.imd.fun

   As of2026-10-19Sample233

   MethodShowHide

   Largest single producing platform in the ERC-8004 sample (api.imd.fun, not in the directory)

   Stage

   erc8004 / Ethereum / registered_by_platform

   Method

   agentURI classified as url; stratified sample: the trailing 30 days split into 12 equal block segments, up to 20 registrations taken from each, applied identically on every chain (tier 2)

   Sample

   233

   As of

   2026-10-19

   Note

   Attribution by agentURI host; a host is strong evidence of the producer, not proof. Sample count, never scaled to the population. A host is where the agent's metadata is served — strong evidence of the producing platform, not proof of it.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='erc8004' and chain='Ethereum' and stage='registered_by_platform' and week='2026-W43';

   Open this number→
8. 16

   Registrations with no URI

   8

   sampled Base registrations that declare no agent URI

   As of2026-10-19Sample127

   MethodShowHide

   Sampled Base registrations that declare no agentURI at all

   Stage

   erc8004 / Base / registered_by_platform

   Method

   agentURI classified as absent; stratified sample: the trailing 30 days split into 12 equal block segments, up to 20 registrations taken from each, applied identically on every chain (tier 1)

   Sample

   127

   As of

   2026-10-19

   Note

   An observation about the registrations, not a parse failure. These registrations declare NO agentURI at all. That is an observation, not a parse failure.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='erc8004' and chain='Base' and stage='registered_by_platform' and week='2026-W43';

   Open this number→

1. 10

   Virtuals token transfers

   Unknown

   sampled Virtuals tokens that moved in the past week

   As of2026-10-19

   MethodShowHide

   Virtuals agent tokens with a transfer in the last 7 days

   Stage

   virtuals / Base / token_active_7d

   Method

   eth_getLogs failed partway: Log response size exceeded. Maximum allowed number of requested blocks is 1000 (tier 3)

   Sample

   n/a

   As of

   2026-10-19

   Note

   Share of the catalogue, denominator stated in the row. Window scan incomplete — unknown, not zero.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='virtuals' and chain='Base' and stage='token_active_7d' and week='2026-W43';

   Open this number→
2. 11

   Real x402 settlements

   291,810

   USDC transfers to catalogued x402 sellers, machine noise excluded

   As of2026-10-19

   MethodShowHide

   x402 settlements to catalogued sellers after the noise filter

   Stage

   x402 / Base / settled_to_catalogued_seller_real

   Method

   the settled_to_catalogued_seller lane minus transfers to sellers where mean transfer < $0.02 AND distinct payers <= 3 over the window (heuristic; thresholds are constants, stated here) (tier 3)

   Sample

   n/a

   As of

   2026-10-19

   Note

   Real vs nominal: mean ≤ $0.02 with ≤ 3 payers is classified as noise and reported as a second figure. A heuristic, not a fact: 592 of 1438 paid sellers (24,993 transfers, $240) were classified as machine noise this window. The nominal row is unchanged and sits beside this one.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='x402' and chain='Base' and stage='settled_to_catalogued_seller_real' and week='2026-W43';

   Open this number→
3. 12

   Paid x402 sellers

   846

   catalogued x402 sellers paid, machine noise excluded

   As of2026-10-19

   MethodShowHide

   Catalogued x402 sellers paid at least once (real)

   Stage

   x402 / Base / seller_paid_real

   Method

   the seller_paid lane minus sellers classified as machine noise by the same rule (tier 3)

   Sample

   n/a

   As of

   2026-10-19

   Note

   Heuristic. A seller with one automated customer is real revenue to that seller; it is excluded here because the question this row answers is breadth of demand, not existence of it.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='x402' and chain='Base' and stage='seller_paid_real' and week='2026-W43';

   Open this number→
4. 14

   Tempo memo transfers

   0

   memo-tagged pathUSD transfers in the sampled blocks

   As of2026-10-19Small sample

   MethodShowHide

   pathUSD TransferWithMemo transfers on Tempo in 7 days

   Stage

   tempo_mpp / Tempo / memo_transfers_7d

   Method

   eth_getLogs for TransferWithMemo on pathUSD 0x20c0000000000000000000000000000000000000, 12/12 stratified segments across the trailing 7 days covering 10.5% of the window's blocks; every event in each segment counted (tier 2)

   Sample

   0

   As of

   2026-10-19

   Note

   Memo transfers are the MPP payment shape; not every memo transfer is an agent payment. MPP can settle over other rails and other TIP-20s, and any transfer may carry a memo. This is a sampled count over one token on one chain — RATES between the three rows are valid, the counts are floors over the sampled blocks.

   Reproduce

   select value, unit, method, sample_size, caveat from agent_metrics_history where venue='tempo_mpp' and chain='Tempo' and stage='memo_transfers_7d' and week='2026-W43';

   Open this number→

1. 17

   Skills with disclosures

   565

   crypto-relevant skills indexed with a disclosure

   As of2026-10-19Sample565

   MethodShowHide

   Crypto-relevant agent skills indexed with a disclosure

   Stage

   skills / crypto_relevant

   Method

   ClawHub + skills.sh public APIs, crypto vocabulary gate, static scan

   Sample

   565

   As of

   2026-10-19

   Note

   A disclosure describes; it is not a safety verdict.

   Reproduce

   select count(*) from skills where crypto_relevant;

   Open this number→
2. 18

   Skills with flags

   41.6%

   of indexed skills carry at least one disclosure flag

   As of2026-10-19Sample565

   MethodShowHide

   Share of those skills carrying at least one disclosure flag

   Stage

   skills / disclosure.flags ≠ []

   Method

   static scan

   Sample

   565

   As of

   2026-10-19

   Note

   A wallet skill that generates keys is doing its job; the flag says so, it does not condemn.

   Reproduce

   select count(*) filter (where jsonb_array_length(disclosure->'flags')>0), count(*) from skills where crypto_relevant;

   Open this number→
3. 19

   Skills handling keys

   163

   skills flagged as generating or handling private keys

   As of2026-10-19Sample565

   MethodShowHide

   Skills that generate or handle private keys

   Stage

   skills / flag generates_or_handles_keys

   Method

   static scan

   Sample

   565

   As of

   2026-10-19

   Reproduce

   select count(*) from skills where disclosure->'flags' ? 'generates_or_handles_keys';

   Open this number→