ageniko

The open trust boundary for AI agents.

One decorator gives every agent action a receipt it can't forge and a gate it can't cross. Open-source, framework-agnostic, 30 seconds to first trace.

Installation

pip install ageniko

Quick Start

from ageniko import instrument, check, approve, agent

# Initialize OTEL tracing
instrument()

# Add receipts to agent functions — proves what it claimed vs what it did
@check
def handle_ticket(ticket):
    result = process(ticket)
    return result

# Add approval gates — blocks before irreversible actions
@approve(when="irreversible")
def delete_account(user_id):
    db.delete_user(user_id)

# Full agent wrapper — tracing + receipts + cost tracking
@agent(name="support-agent", model="gpt-4o")
def run_agent(query):
    return llm.complete(query)

What Ageniko Provides

Self-Host

git clone https://github.com/ageniko-ai/ageniko
cd ageniko
cp .env.example .env
docker compose up

Open http://localhost:3000 for the console.

Pricing

Links

• GitHub
• Documentation
• PyPI

License

Apache 2.0