Show HN: Provena: Open-Source Library for AI Agent Context Governance

**Provena** is an open-source library aimed at AI agent context governance, surfaced on Hacker News as a Show HN under HN AI Agents. The maintainers are actively seeking contributors and report **7 contributors**, **11 “good first issue” labels**, and **17 “help wanted” issues**, all described as well-scoped and backed by code examples.

The project’s technical claim is a gap at the **context input layer**. Existing tooling already covers what agents do (**Microsoft AGT**), what agents say (**Guardrails AI**), and how agents communicate (**NeMo**). Provena targets the data that still flows into the context window without equivalent controls: **retriever results**, **tool outputs**, and **agent messages**.

That matters for engineers building retrieval- and tool-using agents because those inputs shape model behavior before any action, reply, or protocol check runs. Ungoverned retriever hits, tool payloads, and multi-agent traffic can inject noise, policy-violating content, or untrusted material directly into the window that drives the next step.

In market terms, Provena positions itself next to, not instead of, the named stacks: AGT for behavior, Guardrails AI for output speech, NeMo for communication patterns. Its differentiator is governing **inbound** context rather than only outbound actions, spoken content, or inter-agent messaging.

For builders evaluating it now, the practical next steps are the contributor surface and the issue queue: seven people already in the project, eleven good-first-issue tickets, and seventeen help-wanted items with examples. Watch whether those issues mature into clear APIs and policies for filtering or shaping retriever results, tool outputs, and agent messages before they enter the context window.