Trust Scout Fleet

We run
agents
ourselves.

Five Trustprint scout roles run inside the daily crawl. They do not guess success into existence: every run logs whether a candidate was reachable, timed out, failed by contract, or had no callable endpoint at all.

5
Active scout roles
40/5
Probe vs sentinel quota
Daily
Integrated into crawl workflow
QS
Visible in morning health report
scouts · verified run shape
$ npm run scouts
Scout-Base (erc8004/base) ... no_endpoint rows logged
Scout-ETH (erc8004/ethereum) ... no_endpoint rows logged
✓ Scout-Fetch (agentverse/fetch-ai) ... official Almanac endpoint contract-reachable
✓ Scout-HF (huggingface/huggingface) ... inference endpoints reachable
⚠ Scout-Virtuals (virtuals/base) ... current payloads not probeable
Writing results to scout_interactions…
✓ crawl_state updated for scouts/all
$

Meet the Scouts.

Each Scout is a named role in the Trustprint crawl, mapped to a concrete registry and chain. The goal is not to make every source look healthy; the goal is to record what is actually probeable and what is not.

Scout-Base

Keeps a small ERC-8004 Base sentinel sample per run. Current live payloads rarely expose callable endpoints, so rows are logged honestly as no_endpoint without consuming probeable-source quota.
ERC-8004 Base chain current state: limited endpoint coverage
5
sentinel rows / run
Limited endpoint coverage

Scout-ETH

Keeps a small ERC-8004 Ethereum sentinel sample per run. Just like Base, current payloads usually lack stable callable endpoint fields, so the honest result is often no_endpoint.
ERC-8004 Ethereum mainnet current state: limited endpoint coverage
5
sentinel rows / run
Limited endpoint coverage

Scout-Fetch

Discovers official AgentVerse endpoints via Almanac and performs a conservative reachability check. POST-only hosted endpoints are treated as contract-reachable when they return the expected validation response instead of a generic failure.
AgentVerse official Almanac endpoint
40
probe budget / run
Contract-reachable

Scout-HF

Uses stored HuggingFace inference URLs and model endpoints. This is currently the cleanest probe path in production and the strongest example of truly probeable agent targets in the fleet.
HuggingFace inference endpoints
40
probe budget / run
Healthy

Scout-Virtuals

Keeps a small Virtuals Base sentinel sample per run. Current live payloads expose token and market metadata, but not reliable callable agent endpoints, so the honest production state stays not_probeable.
Virtuals current state: not probeable
5
sentinel rows / run
Not probeable today
Why Scout data is different

The data you can't
buy or scrape.

The value of the scout layer is not that every source turns green. The value is that Trustprint records which candidates are truly callable, which fail by contract, and which registries simply do not expose probeable endpoints yet.

5

Registry-specific scout roles

Each role is pinned to a concrete registry/chain combination. That keeps the output explainable: if one path is green and another is yellow, you know exactly which source contract changed.

40/5

Probeability-aware quota

The fleet does not pretend every registry is equally callable. Probeable lanes get more budget; sources with no stable endpoint fields keep small sentinel samples so QS can track the limit without wasting the run.

QS

Visible in the morning report

Scout results are not hidden in ad-hoc logs. They land in scout_interactions, and the daily QS report classifies them as healthy, contract-reachable, not probeable, or missing.

Honest

no_endpoint is still data

A registry without callable endpoints should not be shown as green. Trustprint keeps those rows visible instead of inventing URLs from descriptions or turning source limits into fake uptime.

Coverage Snapshot

What the fleet can
actually verify today.

Current production coverage shape from verified Scout runs. This is not a mock leaderboard; it is the observed reachability state of the scout layer.

#Registry pathStatusNotes
1
HuggingFace
huggingface / huggingface
healthy probeable inference URLs
2
AgentVerse
agentverse / fetch-ai
contract-reachable official Almanac endpoint
3
ERC-8004
base + ethereum scout paths
not_probeable current payloads lack stable endpoints
4
Virtuals
virtuals / base
not_probeable token + market metadata only
5
Next step
broaden probeable coverage before public rankings
roadmap public leaderboard stays secondary

Scout interactions are live in the pipeline today. Start with the public score check for any single address, then move into the API contract if you want to operationalize that signal in a workflow.

Open live score lookup View API contract

Agent Economy Pulse

Weekly Scout summaries — which agents are up, which degraded, and what the behavioral data shows. Every Tuesday.

Free · No spam · Unsubscribe anytime · Privacy Policy