Introducing Tork: AI Governance for Every Framework
February 10, 2026 · 5 min read · Yusuf Jacobs
AI agents are making thousands of decisions per hour — calling tools, accessing databases, sending emails, executing code. But who's watching what they do? Today, we're launching Tork to answer that question.
The Problem
Every major AI framework — LangChain, CrewAI, AutoGen, OpenAI Agents — lets you build powerful autonomous agents. But none of them have built-in governance. There's no standard way to enforce policies on what an agent can do, detect when it handles sensitive data, or prove compliance to auditors.
This isn't a theoretical risk. AI agents with MCP (Model Context Protocol) access can read files, query databases, make API calls, and execute code. One misconfigured tool and your agent is emailing customer SSNs to a third-party API.
What Tork Does
Tork is a governance layer that sits between your AI agent and the outside world. Every action passes through Tork before it executes. In 5 lines of code, you get:
PII Detection & Redaction — 50+ PII types across US, EU, UK, and Australian formats. SSNs, credit cards, healthcare identifiers, biometric data. Detected and redacted before they leave your infrastructure.
Policy Enforcement — Define rules for what your agents can and cannot do. Block dangerous tools, restrict data access, require human approval for sensitive actions.
Cryptographic Receipts — Every governance decision generates an immutable, HMAC hash-chained audit entry, included in the daily on-chain anchor once it reaches Tork. On the on-device package that receipt is a local object: reaching your dashboard and the anchor requires an API key and an SDK version carrying attestation reporting (tork-governance 0.24.0+ on PyPI, 0.11.0+ on npm), and lands as a client attestation. Hand these to your compliance team for HIPAA, GDPR, SOC 2, or PCI-DSS reporting.
MCP Security — Purpose-built for Model Context Protocol. Scan MCP servers, govern tool calls, and enforce policies across Claude, Cursor, and Windsurf agents.
Kill Switch — Instantly halt any agent with a single API call. Because sometimes you need an emergency stop button.
10 Published SDKs, 128 Adapters
We built Tork for every developer, in every language. These are the on-device family — they decide locally. Install with one command:
pip install tork-governance # Python
npm install tork-governance # JavaScript
go get github.com/torkjacobs/tork-go-sdk # Go
cargo add tork-governance # Rust
gem install tork-governance # Ruby
composer require tork/governance # PHP
dotnet add package TorkGovernance # .NET
{:tork_governance, "~> 0.1.0"} # Elixir (mix.exs)
Kotlin: source on GitHub, no verified public registry install yet. 128 framework adapters cover LangChain, CrewAI, AutoGen, Semantic Kernel, DSPy, FastAPI, Express, Gin, Rails, and many more.
The cloud family is a separate package: npm install @torknetwork/sdk, entry class TorkClient. It sends content to Tork and Tork decides server-side — that is the path where the dashboard row records Tork's own decision rather than your attested claim.
Local-First Privacy
Tork's SDKs run entirely on your infrastructure. PII detection, policy enforcement, and receipt generation all happen locally. Your data never leaves your servers unless you choose to sync with Tork Cloud for centralised dashboards and analytics.
This makes Tork suitable for defence, healthcare, financial services, and any environment where data sovereignty matters.
Correction — 3 August 2026
The two paragraphs above describe one product where there are two, and the phrase “unless you choose to sync with Tork Cloud” made syncing sound like a setting on the same package. It is not — it is a different package with a different trust model, and readers who installed the local one and waited for a dashboard were waiting for something that was never going to arrive.
tork-governance (the pip/npm lines above) is the on-device family, entry class Tork. tork-governance makes the governance decision on your machine. PII detection, redaction and the returned verdict are computed on-device. With no API key it makes zero network calls: prompts, completions and detected PII values never leave the machine, and nothing appears in your dashboard.
Supply an API key and the SDK additionally reports a metadata-only attestation of each decision — the action taken, PII type labels and counts, a risk classification, policy labels and a salted fingerprint. It never sends input text, output text or PII values. The decision itself is still made on-device and is never delayed or changed by reporting. Those attestations appear in your dashboard and are included in the daily on-chain anchor — each one recorded as a client attestation (capture_mode=edge, attested_by=client), a claim Tork recorded but did not execute and cannot independently verify. Requires tork-governance 0.24.0+ (PyPI) or 0.11.0+ (npm).
A decision reported by an on-device SDK is recorded as a client attestation (capture_mode=edge, attested_by=client): a claim Tork recorded but did not execute and cannot independently verify. A decision made by @torknetwork/sdk is recorded as capture_mode=cloud, attested_by=tork — Tork made that call itself. Both are equally immutable once anchored; they differ in what is immutable. A client attestation freezes your claim. A server-governed receipt freezes Tork's own decision.
“Receipt generation happens locally” is true, but a local receipt is a local object returned to your code — not a dashboard row. Local governance is free and unlimited — decisions made on your own machine are never metered. Evidence is what is metered: attestations, receipts and anchoring.
Open Source Core
The entire SDK ecosystem is MIT licensed. All 10 published SDKs, all 128 adapters, all PII detection patterns. We believe governance should be accessible to everyone, not locked behind enterprise contracts.
Tork Cloud adds centralised policy management, team dashboards, usage analytics, and enterprise features like SSO and audit exports. Free tier available — no credit card required.
Get Started
Sign up at tork.network/signup and add governance to your first AI agent in under 5 minutes. Check out the documentation for detailed guides, or try the interactive playground to see Tork in action.
We'd love to hear what you're building. Find us on GitHub or email hello@tork.network.
Tork Network Pty Ltd — Sydney, Australia