SDKs
Integrate the authority boundary directly into your stack
TypeScript and Python SDKs, both published and versioned. An autonomous system proposes an action through the SDK, your policy decides, and Parmana enforces before execution proceeds.
Download an SDK
Both are real, published packages, not private previews.
What integrating gets you
Structural enforcement, not a code review
Policy is enforced at the boundary before execution, not left to an autonomous system to interpret and follow on its own.
One SDK surface for every autonomous system
The same client library works whether the caller is an AI agent, a script, or a workflow engine.
Cryptographic proof of what happened
Ed25519 signing, with hybrid ML-DSA signatures available for cryptographic agility and quantum readiness.
Language parity
TypeScript and Python cover the same domain models and error taxonomy, so a team split across both isn't split on behavior.
The integration path
Step 1
Write a policy
Define the rules an action must satisfy: approval limits, vendor allowlists, delegation rules, whatever your existing policy already says.
Step 2
Propose an action through the SDK
Your autonomous system submits a Business Transaction: the action, its target, and the parameters, through the client.
Step 3
Parmana evaluates and enforces
The policy decides, deterministically. An approved action proceeds within your authority. A blocked action never reaches execution.
Step 4
Verify the signed result independently
Every decision produces a signed execution trust record you can verify offline, without trusting Parmana's runtime or database.
Full walkthrough with real captured output: Quickstart
This is a working integration, not a mockup.
Read the docs to see how the authorization boundary works, or talk to us about running a pilot on your own agentic workflow.