# io.github.torkjacobs/tork-governance (npm · @torknetwork/mcp-server)

AI agent governance for MCP: PII detection, policy enforcement, compliance, and kill switch.

- Trust score: 59/100 (low)
- Change this week: +19
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-08-03

## Components

- npm · `@torknetwork/mcp-server`: 59/100 (this document), [markdown](https://verifymcp.io/servers/torkjacobs-tork-governance/torknetwork-mcp-server.md), [page](https://verifymcp.io/servers/torkjacobs-tork-governance/torknetwork-mcp-server)

## Channel facts

- Registry: `npm`
- Package: `@torknetwork/mcp-server`
- Version: `1.5.1`
- Transport: `stdio`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-08-03.

- **Supply Chain Security**: 87/100
  - No malware found by supply-chain analysis.
  - Only part of the dependency tree could be resolved (95 of 99), so this covers what we could see, not the whole tree.
  - No install/post-install scripts declared.
  - Only part of the dependency tree could be resolved (95 of 99), so this covers what we could see, not the whole tree.
- **Provenance & Transparency**: 19/100
  - Repository check failed: the declared repository URL returned HTTP 404.
  - Provenance check failed: no build-provenance attestation is published.
  - Clear OSI-approved license (MIT).
  - Actively maintained (last published 7 days ago).
  - Security-disclosure policy not yet verified: we couldn't inspect the source repository.
- **Schema Quality & AI Usability**: 71/100
  - AI-judged instruction clarity (good).
  - Tool/resource definitions use about 4387 tokens (~91/item across 48 items; 48 tools + 0 resources), lean.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 0/100
  - Stability not yet verified: not enough scan history yet (needs a 30-day window).
- **Tool Coverage**: 100/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 100% of tool parameters carry a description.
- **Capabilities**: 100/100
  - Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.

**Unverified: 1 category.** A category scored 0 because we could not verify it: a data source with nothing on this package, evidence we could not reach, or a check we could not run. We only credit what we can confirm.

## Install

### Claude

```bash
claude mcp add torkjacobs-tork-governance -- npx -y @torknetwork/mcp-server
```

### Codex

```bash
codex mcp add torkjacobs-tork-governance -- npx -y @torknetwork/mcp-server
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "torkjacobs-tork-governance": {
      "type": "local",
      "command": [
        "npx",
        "-y",
        "@torknetwork/mcp-server"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add torkjacobs-tork-governance --command npx --arg -y --arg @torknetwork/mcp-server
```

### Hermes

```yaml
mcp_servers:
  torkjacobs-tork-governance:
    command: "npx"
    args: ["-y", "@torknetwork/mcp-server"]
```

### Other

```json
{
  "mcpServers": {
    "torkjacobs-tork-governance": {
      "command": "npx",
      "args": [
        "-y",
        "@torknetwork/mcp-server"
      ]
    }
  }
}
```

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-08-02 (score 59, +43)

- [security regression] Provenance: unverified → fail
- [security improvement] Install scripts: unverified → pass
- [security improvement] Known CVEs: unverified → partial
- [security improvement] Malware scan: unverified → pass
- [security] Stability: Stability not yet verified: not enough scan history yet (needs a 30-day window).
- [functional regression] Tool coverage: 100 → unverified
- [functional improvement] License: unverified → pass
- [functional improvement] Maintenance: unverified → pass
- [functional improvement] MCP protocol: unverified → pass
- [functional improvement] Schema quality: unverified → good
- [functional improvement] Dependency health: unverified → partial
- [functional] Licence: MIT

### 2026-07-31 (score 16, −2)

- [functional] We updated how we score, so this day's move reflects our rubric, not a change to the server

### 2026-07-28 (score 18, −22)

- [functional regression] Tool coverage: 100 → unverified
- [functional] First check of Schema quality: unverified

### 2026-07-27 (score 40)

First indexed and scored.

## MCP tools (48)

### `tork_scan_pii` (~159 tokens)

Scan text for PII (Personally Identifiable Information) and optionally redact it. Supports regional PII profiles (v1.1) for country-specific and industry-specific detection across 13 countries and 3 industries. Returns found PII types and redacted text.

Input parameters:

- `industry` (string): Industry profile to activate for specialized PII patterns
- `redact` (boolean): Whether to return redacted text (default: true)
- `region` (array): Regional PII profiles to activate (e.g. ['ae', 'in']). Available: au, us, gb, eu, ae, sa, ng, in, jp, cn, kr, br
- `text` (string, required): The text to scan for PII

### `tork_governed_passthrough` (~169 tokens)

Execute a tool call with TORK governance checks. Validates parameters, checks policies, scans for PII with optional regional profiles (v1.1), and logs the call.

Input parameters:

- `agent_id` (string): ID of the agent making the call
- `correlation_id` (string): Optional correlation ID for tracing
- `industry` (string): Industry profile to activate for specialized PII patterns
- `parameters` (object, required): Parameters to pass to the tool
- `region` (array): Regional PII profiles to activate (e.g. ['ae', 'in']). Available: au, us, gb, eu, ae, sa, ng, in, jp, cn, kr, br
- `tool_name` (string, required): Name of the tool to execute

### `tork_check_policy` (~88 tokens)

Check if an action is allowed by the current governance policy without executing it.

Input parameters:

- `action_type` (string, required): Type of action (e.g., 'read', 'write', 'delete', 'execute')
- `agent_id` (string): ID of the agent requesting the action
- `context` (object): Additional context for policy evaluation
- `resource` (string, required): Resource being accessed

### `tork_audit_stats` (~48 tokens)

Get audit statistics for governance events.

Input parameters:

- `agent_id` (string): Filter by specific agent ID
- `window_hours` (number): Time window in hours (default: 24)

### `tork_audit_logs` (~73 tokens)

Query audit logs for governance events.

Input parameters:

- `agent_id` (string): Filter by agent ID
- `event_type` (string): Filter by event type
- `limit` (number): Maximum number of logs to return (default: 50)
- `since` (string): ISO timestamp to filter logs after

### `tork_register_tool` (~169 tokens)

Register a tool with a governance policy. Defines which agents can call it, allowed targets, blocked flags, rate limits, and HITL requirements.

Input parameters:

- `allowed_agents` (array): List of agent IDs allowed to call this tool. Empty array means all agents allowed.
- `allowed_targets` (array): List of allowed target environments (e.g., ['dev', 'staging']). Empty means all allowed.
- `blocked_flags` (array): Additional flags to block beyond defaults (e.g., ['--recursive', '--all'])
- `max_calls_per_minute` (number): Maximum calls allowed per minute (default: 60)
- `requires_hitl` (boolean): Whether this tool requires human-in-the-loop approval
- `tool_name` (string, required): Unique name of the tool to register

### `tork_validate_tool_call` (~112 tokens)

Validate a tool call BEFORE execution. Checks if tool is registered, agent is allowed, target is allowed, parameters don't contain blocked flags, and rate limits are not exceeded.

Input parameters:

- `agent_id` (string, required): ID of the agent making the call
- `parameters` (object, required): Parameters that would be passed to the tool
- `target_environment` (string): Target environment (e.g., 'production', 'staging', 'dev')
- `tool_name` (string, required): Name of the tool to validate

### `tork_tool_call_stats` (~64 tokens)

Get statistics on tool calls including total calls, allowed/blocked counts, and rate limit hits.

Input parameters:

- `include_unregistered` (boolean): Include stats for unregistered tools (default: false)
- `tool_name` (string): Filter by specific tool name (optional)

### `tork_request_approval` (~117 tokens)

Submit an action for human approval. Creates an approval request in the queue. Used when a tool requires HITL or for high-risk actions.

Input parameters:

- `agent_id` (string, required): ID of the agent requesting approval
- `amount` (number): Optional monetary amount (for velocity tracking)
- `parameters` (string, required): JSON string of parameters for the action
- `target` (string, required): Target resource or environment (e.g., 'production', 'user@example.com')
- `tool_name` (string, required): Name of the tool requiring approval

### `tork_approve_action` (~70 tokens)

Human approves a pending action. Checks for reviewer fatigue and enforces cool-down periods.

Input parameters:

- `approval_id` (string, required): ID of the approval request to approve
- `reason` (string): Optional reason for approval
- `reviewer_id` (string, required): ID of the human reviewer approving the action

### `tork_reject_action` (~71 tokens)

Human rejects a pending action. Does not trigger cool-down as rejections indicate vigilance.

Input parameters:

- `approval_id` (string, required): ID of the approval request to reject
- `reason` (string, required): Reason for rejection (required)
- `reviewer_id` (string, required): ID of the human reviewer rejecting the action

### `tork_check_approval` (~43 tokens)

Check the status of an approval request. Returns current status and whether it has expired.

Input parameters:

- `approval_id` (string, required): ID of the approval request to check

### `tork_velocity_alert` (~69 tokens)

Check for slicing attack patterns - detecting many small requests that aggregate to significant impact. Returns velocity stats and alerts.

Input parameters:

- `agent_id` (string): Agent ID to check velocity for (optional, checks all if not provided)
- `target` (string): Target/vendor to check velocity for (optional)

### `tork_dashboard` (~38 tokens)

Get dashboard statistics from Tork Cloud. Shows total tools, agents, policies, pending approvals, and recent activity. Requires TORK_API_KEY.

### `tork_list_tools` (~37 tokens)

List all registered tools from Tork Cloud. Shows tool configurations, rate limits, and HITL requirements. Requires TORK_API_KEY.

### `tork_list_policies` (~36 tokens)

List all governance policies from Tork Cloud. Shows policy names, priorities, and YAML content. Requires TORK_API_KEY.

### `tork_get_policy` (~41 tokens)

Get a specific governance policy by name from Tork Cloud. Requires TORK_API_KEY.

Input parameters:

- `name` (string, required): Name of the policy to retrieve

### `tork_list_approvals` (~52 tokens)

List pending approval requests from Tork Cloud. Filter by status (pending, approved, rejected). Requires TORK_API_KEY.

Input parameters:

- `status` (string): Filter by approval status (default: pending)

### `tork_approve` (~73 tokens)

Approve a pending HITL request in Tork Cloud. Requires reviewer_id. Requires TORK_API_KEY.

Input parameters:

- `approval_id` (string, required): ID of the approval request
- `comment` (string): Optional comment explaining the approval
- `reviewer_id` (string, required): ID of the human reviewer approving the action

### `tork_deny` (~75 tokens)

Deny a pending HITL request in Tork Cloud. Requires reviewer_id and reason. Requires TORK_API_KEY.

Input parameters:

- `approval_id` (string, required): ID of the approval request
- `reason` (string, required): Reason for denial (required)
- `reviewer_id` (string, required): ID of the human reviewer denying the action

### `tork_cloud_audit_logs` (~91 tokens)

Query audit logs from Tork Cloud. Supports filtering by agent, action type, and date. Requires TORK_API_KEY.

Input parameters:

- `action` (string): Filter by action type
- `agent_id` (string): Filter by agent ID
- `limit` (number): Maximum number of logs to return (default: 50)
- `since` (string): ISO timestamp to filter logs after

### `tork_create_webhook` (~109 tokens)

Create a webhook for HITL notifications. Webhooks are called when approvals are requested, approved, or rejected. Requires TORK_API_KEY.

Input parameters:

- `event_types` (array): Event types to subscribe to. Options: approval_requested, approval_approved, approval_rejected, approval_timeout. Defaults to ['approval_requested', 'approval_timeout']
- `name` (string, required): Name for the webhook (e.g., 'Slack Alerts')
- `url` (string, required): URL to receive webhook POST requests

### `tork_list_webhooks` (~26 tokens)

List all configured webhooks for the organization. Requires TORK_API_KEY.

### `tork_delete_webhook` (~37 tokens)

Delete a webhook by ID. Requires TORK_API_KEY.

Input parameters:

- `webhook_id` (string, required): ID of the webhook to delete

### `tork_governance_check` (~173 tokens)

Check if an action is allowed by cloud governance policies. Supports regional PII profiles (v1.1) for country-specific detection. Returns permission status and matched policy. Requires TORK_API_KEY.

Input parameters:

- `agent_id` (string, required): ID of the agent requesting the action
- `industry` (string): Industry profile to activate for specialized PII patterns
- `parameters` (object): Parameters being passed to the tool (optional)
- `region` (array): Regional PII profiles to activate (e.g. ['ae', 'in']). Available: au, us, gb, eu, ae, sa, ng, in, jp, cn, kr, br
- `target` (string): Target environment or resource (optional)
- `tool_name` (string, required): Name of the tool being called

### `tork_memory_snapshot` (~94 tokens)

Create a memory fingerprint/snapshot for an agent. Stores a hash of the current memory state for later verification. Use this after significant memory operations to establish a baseline.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `memory_state` (object, required): The current memory state to snapshot (object or string)
- `metadata` (object): Optional metadata about the snapshot (e.g., session_id, context)

### `tork_memory_verify` (~69 tokens)

Verify current memory against the last snapshot. Detects unauthorized modifications by comparing hashes and identifying changes. Returns drift detection and trust score impact.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `current_memory_state` (object, required): The current memory state to verify (object or string)

### `tork_memory_trust_score` (~170 tokens)

Get or calculate trust score for an agent based on memory integrity. Trust = (consistency × 0.3) + (compliance × 0.3) + (transparency × 0.2) + (predictability × 0.2). Levels: >0.8 trusted, 0.5-0.8 caution, <0.5 untrusted.

Input parameters:

- `action` (string): Action to perform: 'get' retrieves current score, 'calculate' recalculates, 'stats' returns full stats, 'untrusted' lists agents below threshold
- `agent_id` (string, required): Unique identifier for the agent
- `max_trust_level` (string): For 'untrusted' action: maximum trust level to include (default: caution)

### `tork_memory_track_modification` (~173 tokens)

Record a memory modification for audit trail. Track all changes to agent memory including adds, updates, deletes, and clears. Essential for compliance and trust scoring.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `authorized` (boolean): Whether this modification was authorized (default: true)
- `authorized_by` (string): Who/what authorized the modification
- `key_path` (string): Path to the modified key (e.g., 'user.preferences.theme')
- `metadata` (object): Additional metadata about the modification
- `modification_type` (string, required): Type of modification being made
- `new_value` (object): New value (for adds and updates)
- `old_value` (object): Previous value (for updates and deletes)
- `reason` (string): Reason for the modification

### `tork_hitl_check_velocity` (~105 tokens)

Check velocity limits for an agent/approver. Returns whether additional approvals are allowed based on rate limits. Use before processing approvals to prevent bypass attacks.

Input parameters:

- `action_type` (string): Type of action being approved (default: 'default')
- `agent_id` (string, required): Unique identifier for the agent
- `approver_id` (string, required): ID of the human approver
- `value` (number): Optional value/amount of the action for aggregate tracking

### `tork_hitl_create_cooldown` (~108 tokens)

Create a cooldown period for an agent. Prevents approvals during the cooldown period. Use when slicing attacks are detected or for manual security lockdowns.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `approver_id` (string): Optional: Specific approver to cooldown
- `cooldown_type` (string, required): Type of cooldown
- `duration_minutes` (number, required): Duration of the cooldown in minutes
- `reason` (string, required): Reason for the cooldown

### `tork_hitl_check_cooldown` (~67 tokens)

Check if an agent is currently in a cooldown period. Returns cooldown status, type, reason, and remaining time.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `approver_id` (string): Optional: Check cooldown for specific approver

### `tork_hitl_detect_slicing` (~73 tokens)

Detect slicing attacks - patterns of many small requests that bypass individual review thresholds. Checks for rapid-fire approvals, value aggregation, and pattern anomalies.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent to analyze
- `approver_id` (string): Optional: Analyze patterns for specific approver

### `tork_hitl_get_config` (~50 tokens)

Get HITL enforcement configuration including velocity limits, cooldown defaults, and slicing detection thresholds.

Input parameters:

- `agent_id` (string): Optional: Get config for a specific agent (returns agent-specific overrides)

### `tork_mcp_log_call` (~159 tokens)

Log an MCP tool invocation for auditing and compliance. Use this for self-reporting tool calls. Records agent, tool, parameters, result, execution time, and success status.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent making the call
- `error_message` (string): Error message if the call failed
- `execution_time_ms` (number): Execution time in milliseconds
- `result` (object): Result returned by the tool
- `session_id` (string): Optional session identifier for grouping related calls
- `success` (boolean): Whether the call succeeded (default: true)
- `tool_name` (string, required): Name of the tool being invoked
- `tool_parameters` (object): Parameters passed to the tool

### `tork_mcp_get_stats` (~95 tokens)

Get aggregated statistics for MCP tool usage. Shows total calls, success/failure rates, and average execution times per tool.

Input parameters:

- `limit` (number): Maximum number of tool stats to return (default: 100)
- `order_by` (string): Field to order results by (default: total_calls)
- `tool_name` (string): Get stats for a specific tool (optional, returns all if not specified)

### `tork_mcp_get_alerts` (~119 tokens)

Get unacknowledged security and anomaly alerts for MCP tool usage. Alerts include high frequency, unusual patterns, repeated failures, and suspicious parameters.

Input parameters:

- `agent_id` (string): Filter alerts by agent ID (optional)
- `alert_type` (string): Filter by alert type (optional)
- `include_acknowledged` (boolean): Include acknowledged alerts (default: false)
- `limit` (number): Maximum number of alerts to return (default: 50)
- `severity` (string): Filter by severity level (optional)

### `tork_mcp_analyze` (~97 tokens)

Run anomaly detection for an agent's MCP tool usage. Checks for high frequency calls, repeated failures, unusual patterns, suspicious parameters, and new tool usage.

Input parameters:

- `agent_id` (string, required): Agent ID to analyze
- `high_frequency_threshold` (number): Threshold for high frequency detection (default: 100 calls/minute)
- `repeated_failure_threshold` (number): Threshold for repeated failure detection (default: 5 consecutive)

### `tork_supply_chain_register` (~165 tokens)

Register a dependency for an AI agent. Track models, packages, tools, services, APIs, and MCP servers that the agent depends on. Essential for supply chain visibility and health monitoring.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `dependency_type` (string, required): Type of dependency being registered
- `integrity_hash` (string): Integrity hash for verification (optional)
- `metadata` (object): Additional metadata about the dependency
- `name` (string, required): Name of the dependency (e.g., 'gpt-4', 'lodash', 'stripe-api')
- `source` (string): Source registry or provider (e.g., 'npm', 'pypi', 'openai')
- `version` (string): Version string of the dependency

### `tork_supply_chain_deps` (~105 tokens)

Get dependencies for an AI agent. Returns a list of all registered dependencies or a dependency tree structure showing relationships.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `dependency_type` (string): Filter by dependency type (optional)
- `format` (string): Output format: 'list' for flat list, 'tree' for hierarchical view (default: list)
- `include_unverified` (boolean): Include unverified dependencies (default: true)

### `tork_supply_chain_verify` (~96 tokens)

Verify a dependency against trusted sources. Checks if the source is in the trusted registry and validates the integrity hash if provided.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `dependency_name` (string, required): Name of the dependency to verify
- `dependency_type` (string): Type of dependency (optional, for more specific verification)
- `expected_hash` (string): Expected integrity hash to verify against (optional)

### `tork_supply_chain_vulns` (~80 tokens)

Check for known vulnerabilities in agent dependencies. Returns vulnerability matches with severity levels and remediation guidance.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `include_acknowledged` (boolean): Include acknowledged/mitigated vulnerabilities (default: false)
- `severity_filter` (string): Filter by minimum severity level (optional)

### `tork_supply_chain_health` (~66 tokens)

Get supply chain health score and report for an agent. Calculates health based on dependency verification status, vulnerability counts, and trust levels.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `include_report` (boolean): Include detailed health report (default: true)

### `tork_cost_record` (~185 tokens)

Record a cost transaction for an AI agent. Track LLM calls, embeddings, tool usage, API calls, storage, and compute costs. Essential for budget governance and spend tracking.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `input_tokens` (number): Number of input tokens
- `metadata` (object): Additional metadata about the transaction
- `model` (string): Model name (e.g., 'gpt-4', 'claude-3-opus')
- `output_tokens` (number): Number of output tokens
- `provider` (string): Provider name (e.g., 'openai', 'anthropic', 'aws')
- `session_id` (string): Optional session ID for grouping transactions
- `total_cost` (number, required): Total cost of the transaction in USD
- `transaction_type` (string, required): Type of cost transaction

### `tork_cost_check` (~81 tokens)

Check if a proposed spend is allowed before making an API call. Validates against budgets and rate limits. Use this before expensive operations to prevent budget overruns.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `proposed_cost` (number, required): Proposed cost amount in USD
- `proposed_tokens` (number): Proposed token count (optional)

### `tork_cost_summary` (~59 tokens)

Get current spend summary for an agent. Shows total cost, transaction count, top providers/models, and budget status.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent
- `period` (string): Summary period (default: daily)

### `tork_cost_alerts` (~85 tokens)

Get unacknowledged cost alerts for an agent. Alerts include budget warnings, threshold breaches, and anomaly detections.

Input parameters:

- `agent_id` (string): Filter alerts by agent ID (optional)
- `include_acknowledged` (boolean): Include acknowledged alerts (default: false)
- `limit` (number): Maximum number of alerts to return (default: 20)

### `tork_cost_budget_status` (~46 tokens)

Check budget status for an agent. Returns all active budgets with current spend, remaining amount, and percent used.

Input parameters:

- `agent_id` (string, required): Unique identifier for the agent

## Diagnostics

Captured diagnostic sections: Provenance, Dependencies. The full working is on the page: https://verifymcp.io/servers/torkjacobs-tork-governance/torknetwork-mcp-server#diagnostics

## Score history

- 2026-08-03: 59
- 2026-08-02: 59
- 2026-08-01: 16
- 2026-07-31: 16
- 2026-07-30: 18
- 2026-07-28: 18
- 2026-07-27: 40

## Links

- npm package: https://www.npmjs.com/package/@torknetwork/mcp-server
- Socket report: https://socket.dev/npm/package/@torknetwork/mcp-server
- Changelog RSS feed: https://verifymcp.io/servers/torkjacobs-tork-governance/torknetwork-mcp-server/changelog.xml
- Changelog JSON feed: https://verifymcp.io/servers/torkjacobs-tork-governance/torknetwork-mcp-server/changelog.json
- HTML version of this page: https://verifymcp.io/servers/torkjacobs-tork-governance/torknetwork-mcp-server