How It Works

End-to-end flow of intent plan creation, verification, and enforcement.

How It Works

The following diagram shows the end-to-end flow of how the Intent Assurance Plan processes and enforces execution plans:

1. SDK/Agent submits an execution plan

2. Backend validates policies and builds canonical plan

3. CSRG builds deterministic reasoning graph

4. Merkle tree computed over graph nodes

5. Ed25519 token signed with plan_hash

6. Token + step proofs returned to SDK

7. Each tool call verified:
   a. JWT signature check
   b. Backend step verification
   c. CSRG Merkle proof verification

8. If verified → request forwarded to MCP server
   If not → request blocked + audit logged

The CSRG (Canonical Structured Reasoning Graph) ensures cryptographic non-repudiation - once a plan is committed, neither the agent nor any intermediary can execute unauthorized tool calls without detection.

On this page