The full Verify trust corpus, as a licensed bulk dataset

Instead of querying one server at a time, download the full MCP server corpus Verify tracks as a daily snapshot -- including servers Verify hasn't scored or validated yet, with those fields simply null on those rows -- or just what changed since yesterday. Built for registries, agent platforms, security vendors, and data teams who need the whole picture, not one lookup at a time.

Daily snapshot

Every exportable server's current trust score, risk, evidence counts, and capability facts -- Parquet and JSONL.gz.

Daily change feed

Only servers whose trust state materially changed since the last export -- score, risk, evidence, tooling, auth, or transport.

Licensed access

Authenticated, scoped API keys -- same organization/key infrastructure as the Verify Intelligence API.

Request access Download a sample (JSON)

Current feed status

current_snapshot

86,423 rows · 14.6 MB across jsonl_gz/parquet files · generated 2026-08-28 03:16 UTC

change_delta

No material changes in latest delta · generated 2026-08-28 05:44 UTC (0 rows · 52 B)

Two datasets

current_snapshot -- one row per exportable server, every field Verify tracks for trust, risk, evidence, and capabilities, refreshed daily.

change_delta -- one row per materially significant trust change detected since the previous successful export; a missed day self-heals by widening the lookback window rather than silently dropping changes.

Access

Capability	Endpoint
Catalog	`GET /v1/intelligence/data`
List artifacts	`GET /v1/intelligence/data/{dataset}/artifacts`
Download (signed URL)	`POST /v1/intelligence/data/{dataset}/artifacts/{artifact_id}/download`

Downloads are short-lived signed URLs, never a direct file proxy. Access is granted per dataset -- a key scoped to the snapshot feed alone cannot see or list the change feed, and vice versa.

Data dictionary -- current_snapshot

Field	Type	Nullable	Meaning
`schema_version`	string	no	Always "verify.trust_data_feed.snapshot.v1" for this dataset.
`snapshot_as_of`	datetime	no	Timestamp of the export run this row belongs to -- shared across every row in the file.
`registry_identifier`	string	no	Canonical namespace/name identifier.
`namespace`	string	no	Registry namespace.
`name`	string	no	Server name within its namespace.
`title`	string	no	Display title.
`registry_source`	string	no	Which upstream registry Verify discovered this server through.
`remote_url`	string	yes	The server's remote MCP endpoint URL.
`transport_type`	string	yes	e.g. streamable-http, sse, stdio.
`has_oauth`	boolean	no	Whether the server exposes OAuth-protected resource metadata.
`has_dcr`	boolean	no	Whether the server supports OAuth dynamic client registration.
`tool_count`	integer	no	Number of tools the server's most recent validation observed.
`trust_score`	float	yes	Verify's public trust score (0-100).
`risk_level`	string	yes	low / medium / high / critical.
`confidence`	float	yes	Evidence confidence, 0.0-1.0 -- how much live validation evidence backs the score.
`evidence_critical`	integer	no	Count of active critical-severity findings.
`evidence_warning`	integer	no	Count of active warning-severity findings.
`evidence_informational`	integer	no	Count of active informational findings.
`evidence_total`	integer	no	Total active findings across all severities.
`last_validated_at`	string	yes	ISO timestamp of the server's most recent validation run.
`freshness_state`	string	no	fresh / stale, based on Verify's freshness policy.
`deployment_readiness`	string	yes	Verify's production-readiness classification code for this server.
`remote_endpoint_fingerprint`	string	yes	Stable fingerprint of the server's remote endpoint identity.
`server_card_fingerprint`	string	yes	Stable fingerprint of the server's published server-card content.
`safe_for_company_knowledge`	boolean	no	Publisher-set flag for internal-knowledge-base suitability.
`verify_first_tracked_at`	datetime	no	When Verify first ingested this server -- not when it first appeared publicly online.
`scoring_version`	string	yes	Which version of Verify's public scoring algorithm produced trust_score.

Data dictionary -- change_delta

Field	Type	Nullable	Meaning
`schema_version`	string	no	Always "verify.trust_data_feed.change.v1" for this dataset.
`snapshot_as_of`	datetime	no	Timestamp of the export run this row belongs to.
`change_id`	string	no	Unique id for this change record.
`registry_identifier`	string	no	Canonical namespace/name identifier of the affected server.
`detected_at`	datetime	no	When Verify detected this change.
`change_type`	string	no	score_changed / risk_changed / evidence_changed / tooling_changed / auth_changed / transport_changed.
`severity`	string	no	low / medium / high.
`field`	string	no	Dotted path of the specific field that changed, e.g. trust.score.
`previous_value`	any	yes	The field's value before this change.
`current_value`	any	yes	The field's value after this change.
`material`	boolean	no	Always true -- only materially significant changes are exported.
`evidence_reference`	string	yes	Snapshot id tying this change back to the evidence that produced it.
`verify_build`	string	yes	The Verify build that detected this change. Null for rows recorded before this field was wired up -- historical rows are not backfilled.

Commercial-use boundary

This dataset is a licensed commercial product, not a public export -- access requires a scoped, authenticated API key and is granted on request. Redistribution rights are a separate, explicitly grantable scope from read access.

Terms • Verify Intelligence API (per-request access)