kontext-sdk 0.13.0

Proof of compliance Python client for AI agents making agentic stablecoin and fiat payments.

kontext-sdk

Python client for the Kontext trust layer. Cryptographic verifiable intent for org-wide payments.

Thin HTTP client wrapping the Kontext REST API. All compliance logic runs server-side -- this package provides typed Python bindings with sync and async support.

Install

pip install kontext-sdk

Quick start

from kontext import Kontext

ctx = Kontext(api_key="sk_...", project_id="my-agent")

# Log a transaction
ctx.log_transaction(
    tx_hash="0xabc...",
    chain="base",
    amount="5000",
    token="USDC",
    from_address="0xsender...",
    to_address="0xrecipient...",
    agent_id="payment-agent",
)

ctx.flush()

# Trust score
trust = ctx.get_trust_score("payment-agent")
print(f"Trust: {trust.score}/100 ({trust.level})")

Async

from kontext import AsyncKontext

async with AsyncKontext(api_key="sk_...", project_id="my-agent") as ctx:
    await ctx.log(action="transfer", agent_id="agent-1")
    trust = await ctx.get_trust_score("agent-1")

API

Method	Description
log(action, agent_id, ...)	Buffer an action log
log_transaction(tx_hash, chain, ...)	Buffer a transaction log
flush()	Send buffered actions to server
create_task(description, agent_id, ...)	Create human-in-the-loop task
confirm_task(task_id, evidence)	Confirm task with evidence
get_trust_score(agent_id)	Get agent trust score (0-100)
export_audit(format)	Export audit trail (JSON or CSV)
evaluate_anomalies(amount, agent_id, ...)	Evaluate for anomalies
health()	Check API health

TypeScript SDK

The full TypeScript SDK (kontext-sdk on npm) includes auto-instrumentation, pluggable sanctions screening, tamper-evident digest chains, and MCP server mode. See getkontext.com/docs.

License

MIT