Signed audit trails for Agno multi-agent workflows

[jagmarques]

Built a wrapper that cryptographically signs every agent action and tool call in Agno teams, producing a tamper-evident audit trail.

When you have multiple agents delegating tasks, it gets hard to prove after the fact who did what and when. This creates a hash-chained log where each entry is signed with ML-DSA-65.

from agno.agent import Agent
from agno.models.openai import OpenAIChat
import asqav

asqav.init(api_key="sk_...")
auditor = asqav.Agent.create("agno-auditor")
session = auditor.start_session(name="workflow-audit")

agent = Agent(model=OpenAIChat(id="gpt-4o"), name="analyst")
response = agent.run("Analyze Q1 financial data")

for msg in response.messages:
    session.log(f"agent:{msg.role}", metadata={
        "agent": "analyst",
        "content_hash": asqav.hash_content(str(msg.content)),
        "model": "gpt-4o"
    })

Every log entry is hash-chained to the previous one, so you can detect if any entry was inserted, deleted, or modified. Exportable as JSON for compliance reviews.

Useful if you need to demonstrate to auditors exactly what your multi-agent system did - relevant for EU AI Act Article 12.

Repo: https://github.com/Upsonic/Upsonic