MCP Gateway • HITL • Tool GovernanceGovern Every Agent, Every Tool Call
Tork sits between your AI agents and the outside world. Every MCP tool call, every API request, every database query passes through governance before it executes.
MCP Gateway
Purpose-built for Model Context Protocol. Scan MCP servers, govern tool calls, and enforce policies across Claude, Cursor, and Windsurf agents.
Human-in-the-Loop
Require manual approval for high-risk actions. Your agents propose, your team decides. Slack, email, and dashboard integrations.
Kill Switch
Instantly halt any agent with a single API call. Block all tool calls, notify your team, and generate an audit trail — in milliseconds.
Policy Engine
Define JSON rules for what agents can do. Block dangerous tools, restrict data access, and enforce least-privilege across your fleet.
PII Detection
50+ PII types across 13 countries. Detect and redact SSNs, credit cards, Medicare numbers, and more — all running locally on your infrastructure.
Compliance Receipts
Every governance decision generates a cryptographic, hash-chained audit entry. Hand these to auditors for HIPAA, GDPR, SOC 2, or PCI-DSS.
Three Steps to Governed AI
Define Policies
Write human-readable YAML policies.
policies:
- name: pii-protection
action: REDACT
patterns: [email, phone]Wrap Your Calls
Add Tork to your AI pipeline.
from tork import PolicyEngine
engine = PolicyEngine("policy.yaml")
result = engine.evaluate(output)Monitor & Audit
Get dashboards and receipts.
# Every evaluation creates
# a signed receipt:
{
"decision": "ALLOW",
"hmac": "a3f2c1..."
}Install in Seconds
11 SDKs — every major language
$ pip install tork-governance$ npm install tork-governanceWorks with Claude Desktop, Cursor, and any MCP-compatible client
View documentation on GitHub →Configure Your Way
Three ways to manage AI governance. Pick what works for you.
Admin Console
Point and click. Embed our React dashboard directly in your app.
<TorkAdminConsole apiKey="..." />
SDK & API
Full programmatic control. Python, JavaScript, or REST API.
engine = GovernanceEngine(api_key="...")
AI-Assisted
Ask Claude or Cursor to configure governance for you.
"Register a tool that blocks --force flag"
All three methods write to the same database. Changes apply everywhere instantly.
Core Tools for AI Governance
Production-ready compliance without the complexity
MCP Gateway
48 governance tools via Model Context Protocol. Drop-in integration with Claude, Cursor, and MCP clients.
Learn more →Tool Call Governance
Block dangerous flags, enforce allowlists. Rate limiting and target environment restrictions.
Learn more →HITL Systems
Human approval with fatigue detection. Velocity alerts catch slicing attacks.
Learn more →Policy Enforcement
Define rules in YAML, enforce everywhere. Block, warn, or allow based on content.
Learn more →PII Redaction
Designed to detect and redact 50+ PII types. Emails, phones, SSNs, credit cards.
Learn more →Compliance Receipts
HMAC-signed, tamper-evident audit trails. Designed to help demonstrate compliance.
Learn more →Listen Mode
Deploy in observation mode first. Tork observes and records your traffic patterns automatically.
Learn more →Tork significantly reduces AI safety risks through multiple detection layers. No AI safety solution guarantees 100% protection. We recommend using Tork as part of a defense-in-depth strategy. See our Terms of Service for details.
Why Tork?
What sets us apart from the competition
Memory Integrity
UniqueDesigned to monitor agent memory drift. Helps identify when agents may deviate from safety instructions or exhibit unexpected behavioral changes.
Framework Agnostic
UniversalWorks with LangChain, CrewAI, AutoGen, OpenAI, Anthropic, and any custom framework. One governance layer for all.
MCP Native
48 Tools48 governance tools via Model Context Protocol. Drop-in integration with Claude Desktop, Cursor, and any MCP client.
Kill Switch CLI
InstantStop any agent instantly with tork agent kill. Freeze for investigation, resume when safe. Full audit trail.
OpenClaw is powerful.
Tork makes it safe.
Potentially malicious skills exist in the MCP ecosystem. Tork Guardian is designed to scan, flag, and help govern skill interactions — reducing risk for your agents.
Trusted by developers building AI
What Governed AI Looks Like
Real governance scenarios powered by Tork.
“A customer service AI agent was leaking email addresses and phone numbers in its responses. Tork’s PII detection layer caught and redacted the sensitive data before it reached end users.”
PII Detection in Action
Customer Service AI Use Case
“Integrating governance took a single import statement and 3 lines of configuration. Every API call now generates a signed compliance receipt automatically.”
5-Minute Integration
Developer Experience Use Case
“The MCP gateway provides a single control plane to monitor what AI agents are accessing, which tools they’re calling, and whether they’re staying within policy boundaries.”
Full Visibility
MCP Governance Use Case
These scenarios illustrate common governance use cases. Results vary based on configuration and deployment.
The TORKING-X Standard
Beyond tooling — a proposed framework for measuring AI governance quality.
Start Free, Scale with Confidence
Simple pricing for teams of all sizes. All plans include core governance features.
- 5,000 API calls/month
- Basic PII detection
- 1 team member
- Community support
- 7-day log retention
- 50,000 API calls/month
- 120 requests/minute
- Advanced PII detection
- 3 team members
- Email support
- 30-day retention
- 500,000 API calls/month
- 300 requests/minute
- Custom policies
- 10 team members
- Priority support
- SSO (Google/GitHub)
- 2,000,000 API calls/month
- 600 requests/minute
- 25 team members
- Dedicated support
- 180-day retention
- SSO (Google/GitHub)
Enterprise Solutions
Custom governance infrastructure for large organizations. Dedicated support, SLAs, and white-glove onboarding.
Starting at $2,000/mo
See Enterprise Plans- Unlimited API calls & team members
- Dedicated support & SLA
- SSO, SAML & custom domains
- Blockchain-anchored audit trails
- Multi-region deployment options
- On-premise deployment (contact us)
The Agent Crisis
How to Build a Safer, Smarter, and More Prosperous World with Autonomous AI
By Yusuf Jacobs · 94 pages · 21 chapters
A practical guide to AI governance. From PII redaction to compliance receipts, learn how to build governance systems that actually work.
Ready to Govern Your AI?
Join developers building trustworthy AI systems with Tork.
Get AI governance insights weekly — no spam, unsubscribe anytime.