Name: WhiteBoxXAI
Author: AgentaFlow

WhiteBoxXAI MCP

AI observability, explainability and governance: drift, fairness, ISO 42001, risk register.

Grade A Remote or local · Streamable HTTP · stdio API key 197 tools v0.3.2 MIT

No reviews yet by AgentaFlowPyPI Runs on your machine or at its publisher’s address — not through us

94/100

trust score

197

tools declared

local

or remote

—

downloads a day

What it does

AI observability, explainability and governance: drift, fairness, ISO 42001, risk register.

Quickstart

# 1 — add it to Claude Code
claude mcp add --scope user --env 'WHITEBOXXAI_MCP_API_KEY=${WHITEBOXXAI_MCP_API_KEY}' --transport stdio whiteboxxai-mcp '--' uvx whiteboxxai-mcp

# 2 — ask your agent for something these tools do
#   whitebox_agent_workflows_create_and_start, whitebox_agent_workflows_start, whitebox_agent_workflows_complete and 194 more

It needs WHITEBOXXAI_MCP_API_KEY: this reads it from your environment, so export it first.

WhiteBoxXAI costs nothing on mcprush: there is no plan to choose here, no cap to set and nothing mcprush can bill you for.

Publishernot claimed

AgentaFlow1 server

1 category

All 1 server

Collected from a public index. Nobody has claimed this account, so nothing here was written by its author — claim it if it is yours.

Installation

Where are you running it?

Every route below installs the same thing and ends at the same approval screen. This one can run either way — at an address its publisher runs, or as a process under your own user, with your files and your network — and this marketplace is in the path of neither.

Claude Codedeclared by this listing

This is a public server: you run it yourself and this marketplace is not in the path. You need uv (docs.astral.sh/uv): uvx comes with it. Claude Code registers it in one command; --scope user makes it available in every project. Claude Code reads WHITEBOXXAI_MCP_API_KEY (WhiteBoxXAI API key, generated from the dashboard under Settings > API Keys) from your environment each time it starts the server, so the value stays out of its config: export it in your shell profile (macOS, Linux), or run setx WHITEBOXXAI_MCP_API_KEY … and open a new terminal (Windows). The '--' is in quotes so that PowerShell passes it on.

claude mcp add --scope user --env 'WHITEBOXXAI_MCP_API_KEY=${WHITEBOXXAI_MCP_API_KEY}' --transport stdio whiteboxxai-mcp '--' uvx whiteboxxai-mcp

Reconnect, or start a new session, and the tools appear in the model’s tool list.

Or connect to its publisher’s address

Its publisher also runs it at https://api.whiteboxxai.com/mcp/: nothing to start on your machine, and this marketplace is not in the path.

claude mcp add --scope user --transport http whiteboxxai-mcp https://api.whiteboxxai.com/mcp/

Free on mcprush, and it runs from its publisher’s own package: this marketplace is not in the path, so there is no key of ours to hold and no card to add.

What it asks of you

It reaches a system of yours, so it needs your own credential rather than the publisher’s: WHITEBOXXAI_MCP_API_KEY — WhiteBoxXAI API key, generated from the dashboard under Settings > API Keys.. You set it in your own client’s config, on your machine — this marketplace never holds it.

What it writes

One config entry your client uses to start the process locally. A local server runs with your file system and your network, which is why it is priced without metering.

What you approve

197 tools, with what each one reads, writes and reaches shown before you agree — the same list on every route above. Read the tool surface.

Tool surface

What the model actually sees. Descriptions are diffed on every release — see version history.

Only tools that write or reach the network

Read from its source code. A server that builds its tool list when it starts can list more tools than this.

Create a new multi-agent workflow.

Parameter	Type	Required	Notes
name	string	yes	—
framework	string	yes	—
meta_data	object	no	—
tags	array	no	—

Mark a workflow as running. Call after registering agents/tasks and before executing.

Parameter	Type	Required	Notes
workflow_id	string	yes	—
inputs	object	no	—

Mark a workflow completed/failed/cancelled and optionally trigger async analytics (metrics, cost, bottlenecks, timeline).

Parameter	Type	Required	Notes
workflow_id	string	yes	—
status	string	yes	—
outputs	object	no	—
error_message	string	no	—
trigger_analytics	boolean	no	—

List workflows for the organization. Returns {workflows, total, skip, limit}.

Parameter	Type	Required	Notes
status	string	no	—
framework	string	no	—
skip	integer	no	—
limit	integer	no	—

Get a workflow by ID.

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Delete a workflow and all related agents/executions/tasks/interactions.

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Parameter	Type	Required	Notes
workflow_id	string	yes	—
name	string	yes	—
role	string	no	—
agent_type	string	no	—
model_name	string	no	—
llm_provider	string	no	—
llm_config	object	no	—
system_prompt	string	no	—
goal	string	no	—
backstory	string	no	—
tools	array	no	—
capabilities	object	no	—
meta_data	object	no	—

List all agents registered in a workflow.

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Log that an agent started executing. Set parent_execution_id for hierarchical/delegated executions.

Parameter	Type	Required	Notes
workflow_id	string	yes	—
agent_id	string	yes	—
inputs	object	no	—
parent_execution_id	string	no	—
execution_order	integer	no	—

Log agent-to-agent communication. interaction_type: handoff, delegation, query, feedback, or broadcast.

Parameter	Type	Required	Notes
workflow_id	string	yes	—
interaction_type	string	yes	—
from_agent_id	string	no	—
to_agent_id	string	no	—
from_execution_id	string	no	—
to_execution_id	string	no	—
message	string	no	—
meta_data	object	no	—

List all agent-to-agent interactions in a workflow, chronologically.

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Create a workflow task. task_type: research/write/analyze/delegate/review.

Parameter	Type	Required	Notes
workflow_id	string	yes	—
task_name	string	yes	—
description	string	no	—
task_type	string	no	—
expected_output	string	no	—
agent_id	string	no	—
parent_task_id	string	no	—
priority	integer	no	—
input_data	object	no	—
context	object	no	—

Update a task's status. status: pending/running/completed/failed/skipped.

Parameter	Type	Required	Notes
task_id	string	yes	—
status	string	yes	—
output_data	object	no	—
error_message	string	no	—

List all tasks in a workflow, with status, assignments, and results.

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Get workflow analytics (tokens, costs, execution counts, durations).

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Get per-agent cost breakdown (tokens, cost, execution count, avg duration). This may take up to ~30s: it runs a Celery task the backend blocks on server-side before responding.

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Identify workflow bottlenecks: slowest agents, slowest tasks, failing agents. This may take up to ~30s: it runs a Celery task the backend blocks on server-side before responding.

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Get a chronological timeline of workflow events (start/end, agent executions, task updates, interactions). This may take up to ~30s:

Parameter	Type	Required	Notes
workflow_id	string	yes	—

Detect correlations for an alert (temporal, model, feature, and pattern-based relationships to other alerts).

Parameter	Type	Required	Notes
alert_id	string	yes	—
look_back_hours	integer	no	—
min_score	number	no	—

Get clusters of highly correlated alerts -- useful for spotting widespread issues or cascading failures.

Parameter	Type	Required	Notes
min_cluster_size	integer	no	—
min_score	number	no	—
hours_back	integer	no	—

Analyze potential root causes (data drift, model degradation, feature anomalies, system issues) for an alert.

Parameter	Type	Required	Notes
alert_id	string	yes	—
analyze_history	boolean	no	—

Mark a previously computed root cause as verified or rejected.

Parameter	Type	Required	Notes
root_cause_id	integer	yes	—
verified	boolean	yes	—
notes	string	no	—

Find past alerts sharing the same root-cause type -- useful for learning from historical resolutions.

Parameter	Type	Required	Notes
cause_type	string	yes	—
limit	integer	no	—

Compute a weighted priority score (0-100) and level

Parameter	Type	Required	Notes
alert_id	string	yes	—
business_impact_weight	number	no	—
urgency_weight	number	no	—
criticality_weight	number	no	—
history_weight	number	no	—

Manually override a calculated priority. adjustment_reason is required justification for the override.

Parameter	Type	Required	Notes
alert_id	string	yes	—
new_priority_score	number	yes	—
new_priority_level	string	yes	—
adjustment_reason	string	yes	—

Get a breakdown of alerts by priority level.

Parameter	Type	Required	Notes
hours_back	integer	no	—

Get the prioritized alert queue for triage and assignment.

Parameter	Type	Required	Notes
limit	integer	no	—

Detect recurring temporal/frequency/correlation/anomaly patterns in alerts.

Parameter	Type	Required	Notes
hours_back	integer	no	—

Get a specific alert pattern's details.

Parameter	Type	Required	Notes
pattern_id	integer	yes	—

Mark a pattern inactive (resolved or no longer relevant).

Parameter	Type	Required	Notes
pattern_id	integer	yes	—

Apply automated throttling/grouping/snoozing to reduce alert noise. Returns a summary of actions taken and recommendations.

Parameter	Type	Required	Notes
enable_auto_throttling	boolean	no	—
enable_intelligent_grouping	boolean	no	—
enable_auto_snooze	boolean	no	—
similarity_threshold	number	no	—
grouping_window_minutes	integer	no	—
auto_snooze_threshold	integer	no	—

Get recommended alert-throttling rules to reduce fatigue without missing critical issues.

Takes no parameters.

Cluster alerts via ML (kmeans/dbscan/hierarchical). Automatically determines the optimal number of clusters if not specified.

Parameter	Type	Required	Notes
clustering_method	string	no	—
num_clusters	integer	no	—

Group alerts occurring in the same time window with similar

Parameter	Type	Required	Notes
time_window_minutes	integer	no	—
similarity_threshold	number	no	—

Forecast alert volume/severity/type for time_horizon

Parameter	Type	Required	Notes
time_horizon	string	no	—

Detect hour-of-day seasonality in alert volume.

Parameter	Type	Required	Notes
days_back	integer	no	—

Combine real alert-volume time-series data with a real

Parameter	Type	Required	Notes
metric	string	no	—
hours_back	integer	no	—

Forecast future alert volume with confidence intervals.

Parameter	Type	Required	Notes
forecast_hours	integer	no	—

Get the full action timeline (triggered/acknowledged/snoozed/ resolved/escalated/commented) for an alert.

Parameter	Type	Required	Notes
alert_id	string	yes	—

Get comprehensive alert analytics (counts, MTTA/MTTR,

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—

Get alert volume as a time series between start and end

Parameter	Type	Required	Notes
start	string	yes	—
end	string	yes	—
interval	string	no	—
group_by	string	no	—
filters	object	no	—

Get alert generation rate trends (hour-over-hour, day-over-day, rolling averages) for the caller's organization.

Takes no parameters.

Group active alerts by a dimension: "type", "severity", "model", or "rule" -- useful for spotting hotspots.

Parameter	Type	Required	Notes
group_by	string	yes	—

Detect alert-fatigue indicators and get recommendations for

Takes no parameters.

Acknowledge an alert (hits /alert-management/alerts/{id}/acknowledge;

Parameter	Type	Required	Notes
alert_id	string	yes	—
notes	string	no	—

Resolve an alert (hits /alert-management/alerts/{id}/resolve;

Parameter	Type	Required	Notes
alert_id	string	yes	—
resolution_notes	string	no	—

Snooze an alert for snooze_minutes (max 1 week; hits

Parameter	Type	Required	Notes
alert_id	string	yes	—
snooze_minutes	integer	yes	—

Add a comment to an alert (comment_type: note/action/resolution).

Parameter	Type	Required	Notes
alert_id	string	yes	—
comment	string	yes	—
comment_type	string	no	—

List all comments on an alert.

Parameter	Type	Required	Notes
alert_id	string	yes	—

Manually trigger processing of pending escalations (normally a scheduled background task).

Takes no parameters.

Manually trigger reactivation of alerts whose snooze period has expired (normally a scheduled background task).

Takes no parameters.

Create an alert rule with threshold or anomaly-based conditions.

Parameter	Type	Required	Notes
name	string	yes	—
alert_type	string	yes	—
severity	string	yes	—
conditions	array	yes	—
model_id	string	no	—
description	string	no	—
logic_operator	string	no	—
notification_channels	array	no	—
throttle_minutes	integer	no	—
cooldown_minutes	integer	no	—
is_active	boolean	no	—

List alert rules for the caller's organization.

Parameter	Type	Required	Notes
model_id	string	no	—
alert_type	string	no	—
is_active	boolean	no	—
skip	integer	no	—
limit	integer	no	—

Get a specific alert rule by ID.

Parameter	Type	Required	Notes
rule_id	string	yes	—

Update an alert rule. Only non-null arguments are changed.

Parameter	Type	Required	Notes
rule_id	string	yes	—
name	string	no	—
description	string	no	—
severity	string	no	—
is_active	boolean	no	—
conditions	array	no	—
throttle_minutes	integer	no	—
cooldown_minutes	integer	no	—

Delete an alert rule.

Parameter	Type	Required	Notes
rule_id	string	yes	—

Test whether an alert rule would trigger given metric values, without creating a real alert instance. Useful for validating a rule before activation.

Parameter	Type	Required	Notes
rule_id	string	yes	—
metric_values	object	no	—
model_id	string	no	—
historical_values	array	no	—
metadata	object	no	—

List triggered alert instances with optional filters.

Parameter	Type	Required	Notes
rule_id	string	no	—
model_id	string	no	—
alert_type	string	no	—
severity	string	no	—
status	string	no	—
hours	integer	no	—
skip	integer	no	—
limit	integer	no	—

Get a specific triggered alert instance by ID.

Parameter	Type	Required	Notes
alert_id	string	yes	—

Get a triggered alert instance together with its model's latest drift reports and a plain-language triage verdict.

Parameter	Type	Required	Notes
alert_id	string	yes	Triggered alert instance ID
drift_report_limit	integer	no	Number of recent drift reports to include when the alert is associated with a model (default 1)

Acknowledge an active alert instance.

Parameter	Type	Required	Notes
alert_id	string	yes	—
user_id	string	yes	—
notes	string	no	—

Resolve an alert instance.

Parameter	Type	Required	Notes
alert_id	string	yes	—
user_id	string	yes	—
resolution_notes	string	no	—

Snooze an active alert instance for snooze_minutes (max 1 week).

Parameter	Type	Required	Notes
alert_id	string	yes	—
snooze_minutes	integer	yes	—

Get alert statistics for the organization (counts, severity

Parameter	Type	Required	Notes
hours	integer	no	—

Get the organization's ISO/IEC 42001 readiness score and clause breakdown.

Takes no parameters.

Get the organization's open ISO/IEC 42001 compliance gaps.

Parameter	Type	Required	Notes
clause_filter	string	no	Optional ISO/IEC 42001 clause prefix (e.g. '4', '8')
severity_filter	string	no	Optional gap severity filter (e.g. 'high', 'critical')

Get a composite ISO/IEC 42001 readiness scorecard, open gap matrix, and linked AI risk register entries in a single call for the ui://whitebox/iso42001-status-card widget.

Parameter	Type	Required	Notes
clause_filter	string	no	—
severity_filter	string	no	—

List all supported compliance frameworks in the Control Catalog (ISO/IEC 42001, GDPR, CCPA, EU AI Act, NIST AI RMF) with control counts.

Takes no parameters.

Get the full control catalog for one compliance framework.

Parameter	Type	Required	Notes
framework	string	yes	Framework slug (e.g. 'iso_42001', 'eu_ai_act', 'nist_ai_rmf', 'gdpr', 'ccpa')

Run an ad-hoc data drift check for a model (not persisted as a report).

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID
window_size	integer	no	Number of recent predictions to analyze (100-10000)
feature_names	array	no	Optional subset of features to analyze (all if omitted)

Run drift analysis and persist the result as a drift report.

Parameter	Type	Required	Notes
model_id	string	yes	—
window_size	integer	no	—

List drift reports for a model, most recent first.

Parameter	Type	Required	Notes
model_id	string	yes	—
limit	integer	no	—
skip	integer	no	—

Get a specific drift report, including per-feature statistics.

Parameter	Type	Required	Notes
model_id	string	yes	—
report_id	string	yes	—

Get a model's drift trend over time (1-90 days).

Parameter	Type	Required	Notes
model_id	string	yes	—
days	integer	no	—

Generate a SHAP/LIME explanation and block until it's computed.

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID
instance	object	yes	Feature values for the instance to explain, e.g. {"age": 42, "income": 50000}
method	string	no	"shap" or "lime" (uses the model's configured default if omitted)
prediction_id	string	no	Optional ID of the prediction this explanation is for
num_features	integer	no	Number of top features to return (1-100)
num_samples	integer	no	Number of samples the explainer draws (100-50000); higher is slower but more stable
use_cache	boolean	no	Reuse a cached explanation for an identical instance if one exists
total_cfs	integer	no	—
desired_class	string	no	—
features_to_vary	string/array	no	—

Start explanation generation for multiple instances at once (non-blocking).

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID
instances	array	yes	Feature values for each instance to explain (max 100)
method	string	no	"shap" or "lime" (uses the model's configured default if omitted)
num_features	integer	no	Number of top features to return per explanation (1-100)
num_samples	integer	no	Number of samples the explainer draws (100-50000)
parallel	boolean	no	Retained for backward compatibility; not read by the backend
total_cfs	integer	no	—
desired_class	string	no	—
features_to_vary	string/array	no	—

Start real SHAP/LIME explanation generation for a model instance.

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID
instance	object	yes	Feature values for the instance to explain, e.g. {"age": 42, "income": 50000}
method	string	no	"shap" or "lime" (uses the model's configured default if omitted)
prediction_id	string	no	Optional ID of the prediction this explanation is for
num_features	integer	no	Number of top features to return (1-100)
num_samples	integer	no	Number of samples the explainer draws (100-50000); higher is slower but more stable
total_cfs	integer	no	—
desired_class	string	no	—
features_to_vary	string/array	no	—

Get an explanation by ID, including its status and (once ready) results.

Parameter	Type	Required	Notes
explanation_id	string	yes	—

List explanations for a model, optionally filtered by method or status.

Parameter	Type	Required	Notes
model_id	string	yes	—
method	string	no	—
status	string	no	—
limit	integer	no	—
offset	integer	no	—

Get the explanation associated with a prediction, if one exists.

Parameter	Type	Required	Notes
prediction_id	string	yes	—

Get aggregated explanation statistics for a model (counts by method/status, average computation time).

Parameter	Type	Required	Notes
model_id	string	yes	—

Create or update a model's explanation configuration (upsert).

Parameter	Type	Required	Notes
model_id	string	yes	—
enabled	boolean	no	—
default_method	string	no	—
auto_explain	boolean	no	—
cache_enabled	boolean	no	—
cache_ttl_hours	integer	no	—
lime_config	object	no	—
shap_config	object	no	—
visualization_config	object	no	—

Get a model's explanation configuration.

Parameter	Type	Required	Notes
model_id	string	yes	—

Delete a model's explanation configuration.

Parameter	Type	Required	Notes
model_id	string	yes	—

Compare 2-20 explanations against each other.

Parameter	Type	Required	Notes
explanation_ids	array	yes	—
comparison_type	string	no	—

Get visualization data for a single explanation.

Parameter	Type	Required	Notes
explanation_id	string	yes	—
plot_type	string	yes	—
max_features	integer	no	—
config	object	no	—

Get visualization data spanning multiple explanations.

Parameter	Type	Required	Notes
explanation_ids	array	yes	—
plot_type	string	yes	—
feature_index	integer	no	—
interaction_index	integer	no	—
max_features	integer	no	—
highlight_index	integer	no	—

Run a bias/fairness audit on model predictions across demographic groups.

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID
sensitive_attributes	array	yes	Protected attributes to analyze (race, gender, age, ethnicity, religion, disability, sexual_orientation, national_origin, marital_status, or custom)
y_true	array	yes	Ground-truth labels, same length/order as y_pred and group_data values
y_pred	array	yes	Predicted labels
group_data	object	yes	Per-attribute group assignments, e.g. {"gender": ["M", "F", "M", ...]}, one array per attribute in sensitive_attributes, same length/order as y_true/y_pred
y_prob	array	no	Optional predicted probabilities (enables calibration metrics)
metrics_to_compute	array	no	Optional subset of fairness metrics to compute (demographic_parity, equal_opportunity, equalized_odds, disparate_impact, predictive_parity, calibration, false_positive_rate_parity, false_negative_rate

Get full bias audit results by audit_id.

Parameter	Type	Required	Notes
audit_id	string	yes	—

List bias audit summaries, optionally filtered to one model.

Parameter	Type	Required	Notes
model_id	string	no	—
limit	integer	no	—
offset	integer	no	—

Get a model's historical bias/fairness trend (1-365 days).

Parameter	Type	Required	Notes
model_id	string	yes	—
days	integer	no	—

Get history for one specific fairness metric on a model.

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID
metric_type	string	yes	One of demographic_parity, equal_opportunity, equalized_odds, disparate_impact, predictive_parity, calibration, false_positive_rate_parity, false_negative_rate_parity, statistical_parity_difference, a
days	integer	no	Lookback window in days (1-365)

Get the most recent bias audit for a model.

Parameter	Type	Required	Notes
model_id	string	yes	—

List governance review boards for the organization.

Takes no parameters.

List governance review requests, optionally filtered.

Parameter	Type	Required	Notes
board_id	string	no	Restrict to one review board
status	string	no	One of draft, submitted, under_review, approved, denied, withdrawn

Get a specific governance review request dossier (board, submitter, decisions, comments, vote tally, and SHA-256 evidence snapshot hash).

Parameter	Type	Required	Notes
request_id	string	yes	—

Record a board member's vote on a governance review request.

Parameter	Type	Required	Notes
request_id	string	yes	UUID of the governance review request
decision	string	yes	One of 'approve', 'deny', 'abstain', 'request_changes'
evidence_snapshot_hash	string	yes	evidence_snapshot_hash from the whitebox_governance_get_review_request result the reviewer saw
totp_code	string	yes	The reviewer's current authenticator code, or a backup code
justification	string	no	Optional rationale, recorded with the vote
conditions	string	no	Optional conditions of approval

Get the RACI grid: review requests and their Responsible /

Parameter	Type	Required	Notes
board_id	string	no	—

Log a single LLM call with real token/cost/latency data.

Parameter	Type	Required	Notes
provider	string	yes	—
model_name	string	yes	—
prompt	string	yes	—
latency_ms	number	yes	—
model_id	string	no	—
completion	string	no	—
prompt_tokens	integer	no	—
completion_tokens	integer	no	—
finish_reason	string	no	—
temperature	number	no	—
max_tokens	integer	no	—
status	string	no	—
error_message	string	no	—
user_id	string	no	—
session_id	string	no	—
request_id	string	no	—
environment	string	no	—
prompt_version_id	string	no	—

Log multiple LLM calls in one request (max 1000). Each dict is shaped like whitebox_llm_monitoring_log_call's arguments.

Parameter	Type	Required	Notes
logs	array	yes	—

Get a single LLM log by ID.

Parameter	Type	Required	Notes
log_id	string	yes	—

Query LLM logs with filters and pagination. Returns {logs, total, skip, limit}.

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
model_id	string	no	—
provider	string	no	—
model_name	string	no	—
status	string	no	—
user_id	string	no	—
session_id	string	no	—
environment	string	no	—
min_cost	number	no	—
max_cost	number	no	—
min_latency_ms	number	no	—
max_latency_ms	number	no	—
skip	integer	no	—
limit	integer	no	—

Get all LLM logs for a specific session.

Parameter	Type	Required	Notes
session_id	string	yes	—

Get aggregated LLM usage statistics (tokens, cost, latency, by_model).

Parameter	Type	Required	Notes
model_id	string	no	—
provider	string	no	—
environment	string	no	—
start_time	string	no	—
end_time	string	no	—

Get recent LLM logs (requires a model_id owned by the caller's org).

Parameter	Type	Required	Notes
limit	integer	no	—
model_id	string	no	—
provider	string	no	—

Get detailed cost breakdown (platform-wide aggregate).

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
include_user	boolean	no	—
include_environment	boolean	no	—

Get LLM latency/throughput performance metrics.

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
model_id	string	no	—
provider	string	no	—

Get token usage trend over time (granularity: hour/day/week).

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
granularity	string	no	—
model_id	string	no	—

Get cost trend over time (platform-wide; granularity: hour/day/week).

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
granularity	string	no	—
provider	string	no	—

Get detailed, grouped usage statistics (granularity: hour/day/week/month).

Parameter	Type	Required	Notes
start_time	string	yes	—
end_time	string	yes	—
granularity	string	no	—
provider	string	no	—
model_name	string	no	—
model_id	string	no	—
user_id	string	no	—
environment	string	no	—

Check whether costs over the recent period exceed a threshold.

Parameter	Type	Required	Notes
period_minutes	integer	no	—
threshold	number	no	—

Check whether average latency over the recent period exceeds a threshold.

Parameter	Type	Required	Notes
period_minutes	integer	no	—
threshold_ms	number	no	—

Check whether the error rate over the recent period exceeds a threshold.

Parameter	Type	Required	Notes
period_minutes	integer	no	—
threshold_percent	number	no	—

Delete LLM logs older than days. Admin-only on the backend.

Parameter	Type	Required	Notes
days	integer	no	—

Analyze transformer attention patterns for a prompt/completion pair.

Parameter	Type	Required	Notes
prompt	string	yes	—
completion	string	no	—
model_id	string	no	—
layers	array	no	—
aggregate_heads	boolean	no	—
attention_weights	array	no	—
top_k	integer	no	—
llm_log_id	string	no	—

Calculate token-level importance scores for a prompt/completion pair.

Parameter	Type	Required	Notes
prompt	string	yes	—
completion	string	yes	—
model_id	string	no	—
method	string	no	—
top_k	integer	no	—
llm_log_id	string	no	—

Test prompt robustness by perturbing it and measuring output stability.

Parameter	Type	Required	Notes
prompt	string	yes	—
baseline_output	string	yes	—
model_id	string	no	—
num_perturbations	integer	no	—
perturbation_type	string	no	—
llm_log_id	string	no	—

Generate minimal prompt edits that change the model's output, useful for understanding decision boundaries.

Parameter	Type	Required	Notes
prompt	string	yes	—
original_output	string	yes	—
target_change	string	no	—
model_id	string	no	—
max_edits	integer	no	—
num_candidates	integer	no	—
search_strategy	string	no	—
llm_log_id	string	no	—

Assess prompt quality across multiple dimensions and get

Parameter	Type	Required	Notes
prompt	string	yes	—
model_id	string	no	—
check_clarity	boolean	no	—
check_specificity	boolean	no	—
check_completeness	boolean	no	—
check_consistency	boolean	no	—
suggest_improvements	boolean	no	—
llm_log_id	string	no	—
prompt_version_id	string	no	—

Get a specific attention/token-importance/sensitivity/counterfactual explanation by ID.

Parameter	Type	Required	Notes
explanation_id	string	yes	—

Get a specific prompt debug analysis by ID.

Parameter	Type	Required	Notes
analysis_id	string	yes	—

List explanations, optionally filtered by LLM log, type

Parameter	Type	Required	Notes
llm_log_id	string	no	—
explanation_type	string	no	—
model_id	string	no	—
limit	integer	no	—

List prompt analyses, optionally filtered by similar prompt text, model, or minimum quality score (0-1).

Parameter	Type	Required	Notes
prompt	string	no	—
model_id	string	no	—
min_quality_score	number	no	—
limit	integer	no	—

Get aggregated LLM-XAI usage statistics (counts by explanation type, average computation time, prompt quality).

Parameter	Type	Required	Notes
model_id	string	no	—
days	integer	no	—

Get attention-heatmap visualization data for an explanation

Parameter	Type	Required	Notes
explanation_id	string	yes	—

Get token-importance bar-chart visualization data for an explanation.

Parameter	Type	Required	Notes
explanation_id	string	yes	—
top_k	integer	no	—

Run multiple explanation analyses on a single LLM log in one call (more efficient than calling each analysis tool separately).

Parameter	Type	Required	Notes
llm_log_id	string	yes	—
include_attention	boolean	no	—
include_token_importance	boolean	no	—
include_sensitivity	boolean	no	—
include_prompt_debug	boolean	no	—

Create a new metric record for a model.

Parameter	Type	Required	Notes
model_id	string	yes	—
metric_type	string	yes	—
metric_name	string	yes	—
metric_value	number	yes	—
metric_metadata	object	no	—
sample_size	integer	no	—

Create multiple metrics at once (max 100). Each dict in metrics needs metric_type/metric_name/metric_value.

Parameter	Type	Required	Notes
model_id	string	yes	—
metrics	array	yes	—

Calculate classification metrics (accuracy/precision/recall/F1/etc.)

Parameter	Type	Required	Notes
y_true	array	yes	—
y_pred	array	yes	—
y_pred_proba	array	no	—
average	string	no	—
pos_label	integer	no	—

Calculate regression metrics (MAE/MSE/RMSE/R2/etc.) from

Parameter	Type	Required	Notes
y_true	array	yes	—
y_pred	array	yes	—

Get metrics recorded for a model, with optional type/date filters.

Parameter	Type	Required	Notes
model_id	string	yes	—
metric_type	string	no	—
start_date	string	no	—
end_date	string	no	—
skip	integer	no	—
limit	integer	no	—

Get the most recent metric of a specific type for a model.

Parameter	Type	Required	Notes
model_id	string	yes	—
metric_type	string	yes	—

Get time series data for a metric over a required date range.

Parameter	Type	Required	Notes
model_id	string	yes	—
metric_type	string	yes	—
start_date	string	yes	—
end_date	string	yes	—

Get metrics aggregated over period (e.g. "daily", "weekly").

Parameter	Type	Required	Notes
model_id	string	yes	—
period	string	yes	—
metric_type	string	no	—
start_date	string	no	—
end_date	string	no	—
force_recompute	boolean	no	—

Get trend statistics for a metric over lookback_days.

Parameter	Type	Required	Notes
model_id	string	yes	—
metric_type	string	yes	—
lookback_days	integer	no	—

Get rolling-window statistics for a metric over window_days.

Parameter	Type	Required	Notes
model_id	string	yes	—
metric_type	string	yes	—
window_days	integer	no	—

Get a summary of all metrics for a model over the last days.

Parameter	Type	Required	Notes
model_id	string	yes	—
days	integer	no	—

Delete all metrics for a model.

Parameter	Type	Required	Notes
model_id	string	yes	—

Parameter	Type	Required	Notes
name	string	yes	Model name (alphanumeric, spaces, dashes, underscores only)
model_type	string	yes	e.g. "classification", "regression"
features	array	yes	List of input feature names
baseline_metrics	object	yes	Baseline performance metrics, e.g. {"accuracy": 0.94}
framework	string	no	ML framework (sklearn, pytorch, tensorflow, etc.)
version	string	no	Model version (semantic versioning, e.g. "1.0.0")
description	string	no	Human-readable description
target_variable	string	no	Name of the target/label variable
fairness_attributes	array	no	Sensitive attributes to track for bias auditing
tags	array	no	Free-form tags for search/filtering

Get a registered model's full record by its model_id.

Parameter	Type	Required	Notes
model_id	string	yes	—

List registered models, optionally filtered by status/type/tags/search text.

Parameter	Type	Required	Notes
skip	integer	no	Number of results to skip (pagination)
limit	integer	no	Maximum results to return (1-1000)
status_filter	string	no	One of ACTIVE, INACTIVE, DEPRECATED, ARCHIVED
model_type	string	no	e.g. "classification", "regression"
search	string	no	Free-text search over name/description
tags	string	no	Comma-separated list of tags to filter by

Partially update a model's metadata. Only provided fields are changed.

Parameter	Type	Required	Notes
model_id	string	yes	—
description	string	no	—
target_variable	string	no	—
fairness_attributes	array	no	—
tags	array	no	—

Update a model's status: ACTIVE, INACTIVE, DEPRECATED, or ARCHIVED.

Parameter	Type	Required	Notes
model_id	string	yes	—
new_status	string	yes	—

List all versions of a model by name, newest first.

Parameter	Type	Required	Notes
model_name	string	yes	—
skip	integer	no	—
limit	integer	no	—

Get the most recently created version of a model by name.

Parameter	Type	Required	Notes
model_name	string	yes	—

Refresh a model's baseline metrics/profile, e.g. after retraining.

Parameter	Type	Required	Notes
model_id	string	yes	—
baseline_metrics	object	yes	—
baseline_data_hash	string	no	—
baseline_data_count	integer	no	—

Archive a model (soft delete; can be restored later).

Parameter	Type	Required	Notes
model_id	string	yes	—

Restore a previously archived model to active status.

Parameter	Type	Required	Notes
model_id	string	yes	—

Permanently delete a model. This is a hard delete and cannot be undone;

Parameter	Type	Required	Notes
model_id	string	yes	—

Compare two Git-backed model versions, returning architecture changes, parameter deltas, and breaking change flags.

Parameter	Type	Required	Notes
version1_id	string	yes	—
version2_id	string	yes	—

Generate a markdown changelog between two Git-backed model versions.

Parameter	Type	Required	Notes
version1_id	string	yes	—
version2_id	string	yes	—

List Git-backed version sources for a model, newest first.

Parameter	Type	Required	Notes
model_id	string	yes	—
limit	integer	no	—

Get the architecture changes timeline across all versions of a model.

Parameter	Type	Required	Notes
model_id	string	yes	—

Log a single model prediction for monitoring.

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID
input_data	object	yes	Input features as a JSON object
output_data	object	yes	The model's prediction/output as a JSON object
prediction_id	string	no	Optional external prediction ID (generated if omitted)
latency_ms	number	no	Prediction latency in milliseconds
metadata	object	no	Additional free-form metadata

Log up to 1000 predictions in a single batch call.

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID
predictions	array	yes	List of objects, each with "input_data" and "output_data" keys (and optionally "prediction_id", "latency_ms", "metadata")

Get a single logged prediction by its prediction_id.

Parameter	Type	Required	Notes
prediction_id	string	yes	—

Query logged predictions for a model, optionally within a time range.

Parameter	Type	Required	Notes
model_id	string	yes	Registered model's ID (required)
start_time	string	no	ISO-8601 timestamp; only predictions after this time
end_time	string	no	ISO-8601 timestamp; only predictions before this time
limit	integer	no	Maximum results to return (1-1000)
offset	integer	no	Number of results to skip

Get prediction volume/latency statistics for a model.

Parameter	Type	Required	Notes
model_id	string	yes	—
start_time	string	no	—
end_time	string	no	—

Get the most recent predictions for a model (max 100).

Parameter	Type	Required	Notes
model_id	string	yes	—
limit	integer	no	—

List all prompt governance templates for the authenticated tenant.

Takes no parameters.

Get details of a specific prompt template by ID.

Parameter	Type	Required	Notes
template_id	string	yes	—

Create a new prompt template container for versioning.

Parameter	Type	Required	Notes
name	string	yes	—
description	string	no	—
tags	object	no	—

List all append-only governed versions for a prompt template.

Parameter	Type	Required	Notes
template_id	string	yes	—

Get details, lifecycle stage, and verification status of a prompt version.

Parameter	Type	Required	Notes
template_id	string	yes	—
version_id	string	yes	—

Create a new governed prompt version in 'design' lifecycle stage.

Parameter	Type	Required	Notes
template_id	string	yes	—
user_message_template	string	yes	—
system_message	string	no	—
model_name	string	no	—
model_id	string	no	—
temperature	number	no	—
tool_schemas	object	no	—
version_tag	string	no	—
commit_hash	string	no	—

Advance a prompt version through the lifecycle stages (design -> verify -> deploy -> retired).

Parameter	Type	Required	Notes
template_id	string	yes	—
version_id	string	yes	—
target_stage	string	yes	—
notes	string	no	—
verification_run_id	string	no	—

Promote a prompt version to active production deployment.

Parameter	Type	Required	Notes
template_id	string	yes	—
version_id	string	yes	—
deployment_notes	string	no	—
force	boolean	no	—

Emergency rollback to a target version or the most recently verified version.

Parameter	Type	Required	Notes
template_id	string	yes	—
target_version_id	string	no	—
reason	string	no	—

Compute prompt drift analysis comparing production completions with baseline expectations.

Parameter	Type	Required	Notes
template_id	string	yes	—
version_id	string	yes	—
baseline_dataset_id	string	no	—
sample_size	integer	no	—

List historical prompt drift reports for a prompt version.

Parameter	Type	Required	Notes
template_id	string	yes	—
version_id	string	yes	—

Get a specific prompt drift report with metrics, divergence, and auditor verdict.

Parameter	Type	Required	Notes
template_id	string	yes	—
version_id	string	yes	—
report_id	string	yes	—

Get security incidents and alerts associated with a prompt version.

Parameter	Type	Required	Notes
template_id	string	yes	—
version_id	string	yes	—

Log a retrieval operation, with real quality metrics computed server-side when ground_truth_ids is supplied.

Parameter	Type	Required	Notes
query	string	yes	—
results	array	yes	—
top_k	integer	yes	—
retrieval_method	string	yes	—
query_type	string	no	—
retrieval_params	object	no	—
retrieval_latency_ms	number	no	—
answer	string	no	—
answer_latency_ms	number	no	—
ground_truth_ids	array	no	—
llm_log_id	string	no	—
model_id	string	no	—
user_id	string	no	—
session_id	string	no	—
environment	string	no	—
metadata	object	no	—

Get a specific retrieval log by ID.

Parameter	Type	Required	Notes
log_id	string	yes	—

Query retrieval logs for a model. model_id is required (owned by the caller's organization) until organization-wide RAG log scoping ships.

Parameter	Type	Required	Notes
model_id	string	yes	—
start_time	string	no	—
end_time	string	no	—
user_id	string	no	—
session_id	string	no	—
retrieval_method	string	no	—
environment	string	no	—
limit	integer	no	—
offset	integer	no	—

Get aggregated RAG statistics (precision/recall/MRR/NDCG/relevance/latency).

Parameter	Type	Required	Notes
model_id	string	no	—
environment	string	no	—
start_time	string	no	—
end_time	string	no	—

Get RAG quality metrics trend over time (granularity: hour/day/week).

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
granularity	string	no	—
model_id	string	no	—
environment	string	no	—

Create a RAG evaluation, aggregating already-logged retrievals for each query. Queries with no matching logged retrieval are reported in detailed_results.queries_without_data, not silently scored as zero.

Parameter	Type	Required	Notes
name	string	yes	—
queries	array	yes	—
description	string	no	—
evaluation_type	string	no	—
model_id	string	no	—
system_version	string	no	—
ground_truth	object	no	—
environment	string	no	—
metadata	object	no	—

Get a specific RAG evaluation by ID.

Parameter	Type	Required	Notes
evaluation_id	string	yes	—

List RAG evaluations with optional filters.

Parameter	Type	Required	Notes
model_id	string	no	—
evaluation_type	string	no	—
environment	string	no	—
limit	integer	no	—
offset	integer	no	—

Get Precision@K metrics for the given K values (default [1, 3, 5, 10]).

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
model_id	string	no	—
environment	string	no	—
k_values	array	no	—

Get average relevance metrics (context/answer/faithfulness).

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
model_id	string	no	—

List AI risk register entries, optionally filtered.

Parameter	Type	Required	Notes
status	string	no	One of identified, assessed, mitigation_planned, mitigated, closed
severity	string	no	One of low, medium, high, critical
category	string	no	Risk category (org-specific taxonomy, e.g. bias, drift, security)
model_id	string	no	Restrict to risks linked to one registered model
limit	integer	no	Page size
skip	integer	no	Page offset

Get a single AI risk register entry by ID.

Parameter	Type	Required	Notes
entry_id	string	yes	—

Get the likelihood x impact heat map and top open risks across the organization's risk register.

Takes no parameters.

Analyze content for toxicity, PII, and harmful content.

Parameter	Type	Required	Notes
content	string	yes	—
content_type	string	no	—
check_toxicity	boolean	no	—
check_pii	boolean	no	—
check_harmful	boolean	no	—
language	string	no	—
llm_log_id	string	no	—
model_id	string	no	—
user_id	string	no	—
session_id	string	no	—
environment	string	no	—

Analyze multiple content items in one request (max 100). Items that fail analysis are silently skipped, not raised.

Parameter	Type	Required	Notes
contents	array	yes	—
content_type	string	no	—
check_toxicity	boolean	no	—
check_pii	boolean	no	—
check_harmful	boolean	no	—
language	string	no	—
model_id	string	no	—
user_id	string	no	—
environment	string	no	—

Get a specific safety score by ID.

Parameter	Type	Required	Notes
score_id	string	yes	—

Query safety analysis scores with filters.

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
model_id	string	no	—
safety_status	string	no	—
contains_pii	boolean	no	—
environment	string	no	—
limit	integer	no	—
offset	integer	no	—

Get aggregated safety statistics (per-category toxicity averages, flagged categories, PII types).

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
model_id	string	no	—
environment	string	no	—

Get content-safety trend over time (granularity: hour/day/week).

Parameter	Type	Required	Notes
start_time	string	no	—
end_time	string	no	—
granularity	string	no	—
model_id	string	no	—
environment	string	no	—

Create a safety threshold configuration controlling when content is flagged as unsafe. action_on_violation: "warning", "block", or "alert".

Parameter	Type	Required	Notes
name	string	yes	—
model_id	string	no	—
description	string	no	—
toxicity_threshold	number	no	—
severe_toxicity_threshold	number	no	—
identity_attack_threshold	number	no	—
insult_threshold	number	no	—
profanity_threshold	number	no	—
threat_threshold	number	no	—
sexually_explicit_threshold	number	no	—
block_pii	boolean	no	—
allowed_pii_types	array	no	—
max_pii_count	integer	no	—
action_on_violation	string	no	—
send_alerts	boolean	no	—

List safety thresholds for the caller's organization.

Parameter	Type	Required	Notes
model_id	string	no	—
is_active	boolean	no	—

Get a specific safety threshold by ID.

Parameter	Type	Required	Notes
threshold_id	string	yes	—

Update a safety threshold configuration. Only non-null arguments are changed.

Parameter	Type	Required	Notes
threshold_id	string	yes	—
name	string	no	—
description	string	no	—
is_active	boolean	no	—
toxicity_threshold	number	no	—
block_pii	boolean	no	—
action_on_violation	string	no	—
send_alerts	boolean	no	—

Delete a safety threshold configuration.

Parameter	Type	Required	Notes
threshold_id	string	yes	—

Free

$0 forever

Every tool, no call limit
No card, and mcprush charges nothing for it
Source published under MIT
Runs on your machine — nothing of it reaches our gateway
Nothing to cap, because mcprush bills nothing

Nothing here is billable: mcprush charges nothing to install WhiteBoxXAI or to call it, at any volume.

ATrust score 94 / 100

Release 0.3.2 · scanned 28 Oct 2026

Two independent axes, because powerful and malicious are different questions. The grade is threat only. The capability level is blast radius, and it is never a penalty on the grade — it is priced as one subtract-only term in the score, where you can see it.

Threat score100

Client score94

Findings1 threat · 1 capability

Capabilityhigh

Coveragesource

Verificationsource

Surface digest2df20984938cb55fb3e4b855c49c1eebce2d9904f072a679312ee2897acdb235

Score vectorthreat 100 → client 94

Term	Level	What it prices	Points
capability-exposure	high	capability blast radius (high) — client exposure if the model is manipulated	−6
verification-discount	source	publisher verification (provenance) — cryptographic build provenance ties the artifact to its source	−0
coverage-honesty	source	inspection depth (source) — how much of the target the scan could see	−0

score = clamp(round(threatScore − E_cap − E_ver − E_cov), 0, 100). No model, no opinion, no network call at scoring time: the same methodology version against the same bytes always produces this number, and every line above is a field of the report rather than a summary of it.

What the scan could actually read

A grade is only as meaningful as its coverage, so the scanner publishes its own depth before it publishes its result.

Tool surfaceread

Implementation sourceread

Package metadataread

Live transportnot available

Tools were statically extracted from the published source (197 recovered), not enumerated from a running server. Tool-poisoning, Unicode-smuggling, capability and toxic-flow analysis ran on this inferred surface, but a mis-parsed registration could be missed or mis-attributed, so tool-derived findings are capped below “confirmed”. To grade the real runtime surface, scan the running server: --command "npx -y <package>".

Capability — what it could do if the model were manipulated

Tags derived from each tool’s schema and the implementation, not from what the tool calls itself. high is the level these add up to.

Tool	Capability tags	Why the tag was assigned
whitebox_agent_workflows_create_and_start	no tags
whitebox_agent_workflows_start	no tags
whitebox_agent_workflows_complete	no tags
whitebox_agent_workflows_list	no tags
whitebox_agent_workflows_get	no tags
whitebox_agent_workflows_delete	no tags
whitebox_agent_workflows_register_agent	no tags
whitebox_agent_workflows_list_agents	no tags
whitebox_agent_workflows_create_execution	no tags
whitebox_agent_workflows_log_interaction	no tags
whitebox_agent_workflows_list_interactions	no tags
whitebox_agent_workflows_create_task	no tags
whitebox_agent_workflows_update_task_status	no tags
whitebox_agent_workflows_list_tasks	no tags
whitebox_agent_workflows_analytics	no tags
whitebox_agent_workflows_cost_breakdown	no tags
whitebox_agent_workflows_bottlenecks	no tags
whitebox_agent_workflows_timeline	no tags
whitebox_alert_intelligence_detect_correlations	no tags
whitebox_alert_intelligence_get_correlation_clusters	no tags
whitebox_alert_intelligence_analyze_root_cause	no tags
whitebox_alert_intelligence_verify_root_cause	no tags
whitebox_alert_intelligence_get_similar_incidents	no tags
whitebox_alert_intelligence_calculate_priority	no tags
whitebox_alert_intelligence_adjust_priority	no tags
whitebox_alert_intelligence_get_priority_distribution	no tags
whitebox_alert_intelligence_get_priority_queue	no tags
whitebox_alert_intelligence_detect_patterns	no tags
whitebox_alert_intelligence_get_pattern	no tags
whitebox_alert_intelligence_deactivate_pattern	no tags
whitebox_alert_intelligence_apply_fatigue_reduction	no tags
whitebox_alert_intelligence_get_throttle_recommendations	no tags
whitebox_alert_intelligence_ml_grouping	no tags
whitebox_alert_intelligence_temporal_grouping	no tags
whitebox_alert_intelligence_predict_trends	no tags
whitebox_alert_intelligence_detect_seasonality	no tags
whitebox_alert_intelligence_enhanced_trend_analysis	no tags
whitebox_alert_intelligence_forecast_volume	no tags
whitebox_alert_management_get_history	no tags
whitebox_alert_management_get_analytics	no tags
whitebox_alert_management_get_trend	no tags
whitebox_alert_management_get_velocity	no tags
whitebox_alert_management_get_grouping	no tags
whitebox_alert_management_detect_fatigue	no tags
whitebox_alert_management_acknowledge	no tags
whitebox_alert_management_resolve	no tags
whitebox_alert_management_snooze	no tags
whitebox_alert_management_create_comment	external-sink	external-sink: keyword "create_comment"
whitebox_alert_management_get_comments	no tags
whitebox_alert_management_process_pending_escalations	no tags
whitebox_alert_management_unsnooze_expired_alerts	no tags
whitebox_alerts_create_rule	no tags
whitebox_alerts_list_rules	no tags
whitebox_alerts_get_rule	no tags
whitebox_alerts_update_rule	no tags
whitebox_alerts_delete_rule	no tags
whitebox_alerts_evaluate_rule	no tags
whitebox_alerts_list_instances	no tags
whitebox_alerts_acknowledge	no tags
whitebox_alerts_resolve	no tags
whitebox_alerts_snooze	no tags
whitebox_alerts_statistics	no tags
whitebox_alerts_get_instance	no tags
whitebox_alerts_get_triage_context	no tags
whitebox_compliance_list_frameworks	no tags
whitebox_compliance_get_framework_controls	no tags
whitebox_compliance_iso42001_score	no tags
whitebox_compliance_iso42001_gaps	no tags
whitebox_compliance_iso42001_status	no tags
whitebox_drift_detect	no tags
whitebox_drift_create_report	no tags
whitebox_drift_list_reports	no tags
whitebox_drift_trend	no tags
whitebox_drift_get_report	no tags
whitebox_explanations_generate_bulk	no tags
whitebox_explanations_generate_async	no tags
whitebox_explanations_list_by_model	no tags
whitebox_explanations_stats	no tags
whitebox_explanations_set_config	no tags
whitebox_explanations_get_config	no tags
whitebox_explanations_delete_config	no tags
whitebox_explanations_compare	no tags
whitebox_explanations_visualize	no tags
whitebox_explanations_visualize_multi	no tags
whitebox_explanations_generate	no tags
whitebox_explanations_get	no tags
whitebox_explanations_get_by_prediction	no tags
whitebox_fairness_audit	no tags
whitebox_fairness_get_audit	no tags
whitebox_fairness_list_audits	no tags
whitebox_fairness_bias_history	no tags
whitebox_fairness_metric_history	no tags
whitebox_fairness_latest_audit	no tags
whitebox_governance_list_boards	no tags
whitebox_governance_raci_grid	no tags
whitebox_governance_list_review_requests	no tags
whitebox_governance_get_review_request	no tags
whitebox_governance_submit_decision	no tags
whitebox_llm_monitoring_log_call	no tags
whitebox_llm_monitoring_log_calls_batch	no tags
whitebox_llm_monitoring_get_log	no tags
whitebox_llm_monitoring_query_logs	no tags
whitebox_llm_monitoring_session_logs	no tags
whitebox_llm_monitoring_usage_stats	no tags
whitebox_llm_monitoring_recent	no tags
whitebox_llm_monitoring_cost_breakdown	no tags
whitebox_llm_monitoring_performance	no tags
whitebox_llm_monitoring_trends_tokens	sensitive-source	sensitive-source: keyword "get_token"
whitebox_llm_monitoring_trends_costs	no tags
whitebox_llm_monitoring_usage_stats_detailed	no tags
whitebox_llm_monitoring_cost_threshold_alert	no tags
whitebox_llm_monitoring_latency_threshold_alert	no tags
whitebox_llm_monitoring_error_rate_alert	no tags
whitebox_llm_monitoring_cleanup_logs	no tags
whitebox_llm_xai_attention	no tags
whitebox_llm_xai_token_importance	no tags
whitebox_llm_xai_prompt_sensitivity	no tags
whitebox_llm_xai_counterfactuals	no tags
whitebox_llm_xai_debug_prompt	no tags
whitebox_llm_xai_get_explanation	no tags
whitebox_llm_xai_get_prompt_analysis	no tags
whitebox_llm_xai_list_explanations	no tags
whitebox_llm_xai_list_prompt_analyses	no tags
whitebox_llm_xai_stats	no tags
whitebox_llm_xai_visualize_attention	no tags
whitebox_llm_xai_visualize_token_importance	sensitive-source	sensitive-source: keyword "get_token"
whitebox_llm_xai_batch_analyze	no tags
whitebox_metrics_create	no tags
whitebox_metrics_create_batch	no tags
whitebox_metrics_calculate_classification	no tags
whitebox_metrics_calculate_regression	no tags
whitebox_metrics_get_model_metrics	no tags
whitebox_metrics_latest	no tags
whitebox_metrics_timeseries	no tags
whitebox_metrics_aggregate	no tags
whitebox_metrics_trend	no tags
whitebox_metrics_rolling	no tags
whitebox_metrics_summary	no tags
whitebox_metrics_delete	no tags
whitebox_models_register	no tags
whitebox_models_get	no tags
whitebox_models_list	no tags
whitebox_models_update	no tags
whitebox_models_update_status	no tags
whitebox_models_versions	no tags
whitebox_models_latest	no tags
whitebox_models_update_baseline	no tags
whitebox_models_archive	no tags
whitebox_models_restore	no tags
whitebox_models_delete	no tags
whitebox_models_list_git_versions	no tags
whitebox_models_compare_versions	no tags
whitebox_models_version_changelog	no tags
whitebox_models_architecture_timeline	no tags
whitebox_predictions_log	no tags
whitebox_predictions_log_batch	no tags
whitebox_predictions_get	no tags
whitebox_predictions_query	no tags
whitebox_predictions_stats	no tags
whitebox_predictions_recent	no tags
whitebox_prompts_list_templates	no tags
whitebox_prompts_get_template	no tags
whitebox_prompts_create_template	no tags
whitebox_prompts_list_versions	no tags
whitebox_prompts_get_version	no tags
whitebox_prompts_create_version	no tags
whitebox_prompts_advance_stage	no tags
whitebox_prompts_deploy	no tags
whitebox_prompts_rollback	no tags
whitebox_prompts_compute_drift	no tags
whitebox_prompts_list_drift_reports	no tags
whitebox_prompts_get_drift_report	no tags
whitebox_prompts_get_incidents	no tags
whitebox_rag_log_retrieval	no tags
whitebox_rag_get_retrieval	no tags
whitebox_rag_list_retrievals	no tags
whitebox_rag_stats	no tags
whitebox_rag_trends	no tags
whitebox_rag_create_evaluation	no tags
whitebox_rag_get_evaluation	no tags
whitebox_rag_list_evaluations	no tags
whitebox_rag_metrics_precision	no tags
whitebox_rag_metrics_relevance	no tags
whitebox_risk_register_list	no tags
whitebox_risk_register_get	no tags
whitebox_risk_register_portfolio	no tags
whitebox_safety_analyze	no tags
whitebox_safety_analyze_batch	no tags
whitebox_safety_get_score	no tags
whitebox_safety_get_scores	no tags
whitebox_safety_stats	no tags
whitebox_safety_trends	no tags
whitebox_safety_create_threshold	no tags
whitebox_safety_list_thresholds	no tags
whitebox_safety_get_threshold	no tags
whitebox_safety_update_threshold	no tags
whitebox_safety_delete_threshold	no tags

Toxic-flow graph

The lethal trifecta, checked as a graph rather than as a checklist: untrusted input, a sensitive source and an external sink have to meet before there is a path worth worrying about.

whitebox_llm_monitoring_trends_tokens → whitebox_alert_management_create_comment

A sensitive-source → external-sink chain exists: whitebox_llm_monitoring_trends_tokens → whitebox_alert_management_create_comment.

Supply chain and provenance

Integrityfirst-seen

Surface digest2df20984938cb55f

Scanned2026-10-28

This is the first scan of this surface here, so there is nothing yet to compare it against.

Every result on this tab comes from one deterministic pass over the published package — offline, rule by rule, and auditable line by line above. Same methodology version, same bytes, same score.

WhiteBoxXAI MCP

What it does

Quickstart

Where are you running it?

Tool surface

What counts against your monthly calls

What the scan could actually read

Capability — what it could do if the model were manipulated

Toxic-flow graph

Supply chain and provenance

Release history

See also