What’s newRelease 11 · Nov 16, 2026Bring Ollive into your own productRead the changelog

Python SDK

The Ollive Python SDK.
Your agents’ record, in your product.

Put Ollive’s findings, evidence and trends wherever your team and your customers already look. A few lines of Python give you everything you see in Ollive, fully typed and ready for production.

Install
pip install ollive
Version
0.1.0
Requires
Python 3.10 or later
Make your first call
An agent page in your product, before and with the Ollive SDK. With the SDK, the same page shows how often the agent meets its rules, the findings behind it with the law they come from, and a compliance result on every conversation.

What you can build

Ollive checks every conversation your agents have against the laws, frameworks and contracts they’re bound by. The SDK puts that record in your hands.

A few lines is all it takes

Every high-severity rule your agents didn’t meet, and how often, with one call:

findings.py

from ollive import Ollive, Severity client = Ollive()  # reads OLLIVE_API_KEY for finding in client.findings.list(severities=[Severity.CRITICAL, Severity.HIGH]):    print(finding.rule.title, finding.in_time_range.run_count, "runs")

Or let your coding agent build it

Describe what you want in your own words, and your coding agent writes the integration. Every method and field is typed and documented, so it gets the calls right the first time, and you review the change like any other.

  • PromptAdd a compliance panel to our agent page with the ollive Python SDK: the share of rules met, with the change on last week, and the top findings with their citations.
  • PromptWrite a job that emails our leadership a weekly AI risk summary from Ollive: findings this week against last week, where the risk sits by rule set, and what to look at first.
  • PromptEvery 15 minutes, check Ollive for new critical findings and send each one to our on-call webhook with the rule, the agent and its citations.

Connect the Ollive MCP server to the same agent, and it can check your real agents and findings while it builds.

Built for production

  • Typed, end to end

    Every request and response has type hints, and every fixed value is an enum. Your editor completes it, and your type checker catches a typo before it ships.

  • Sync or async

    Ollive and AsyncOllive share every method. Use the one your service already runs on.

  • Pages itself

    List methods hand you an iterator. Loop over it, and the SDK fetches the next page as you need it.

  • Retries for you

    Dropped connections, timeouts, rate limits and server errors are retried with backoff, honouring Retry-After.

  • Errors you can act on

    One exception family, with a stable code, the field at fault and a request ID to send support.

  • Ready for what’s next

    New fields and values from Ollive never break the version you’ve installed. Upgrade when it suits you.

See the production guide.

← Docs homeQuickstart →