Find the evidence
Query the catalog and retrieve identified aggregates through the project's interfaces.
Open-source prototype
pgopcore is an open-source prototype for querying registered evidence, checking its integrity and limited consistency, and generating traceable reports through a CLI, Python API and MCP.
Open source · Apache-2.0 · In development
The starting question
In work with agents, a result can become separated from its sources. Reconstructing what was consulted, what was checked and what remains unknown makes review harder.
pgopcore explores how to keep identified evidence and its limits alongside a report, so another person can review them.
It could serve teams developing or reviewing agents. Its usefulness for teams still needs to be validated with users.
The main release
The prototype available on main works with JSON aggregates that are already registered. These are its documented capabilities.
Query the catalog and retrieve identified aggregates through the project's interfaces.
Check the aggregate's hashes and limited consistency, with a receipt for the outcome of that check.
Generate reports and retrieve their artifacts and receipts, including through paginated reads.
An essential limit. Checking integrity does not establish the truth of a declaration or repeat the tests described in a source.
A concrete distinction
A claim about a function's structure and a claim about its behavior need different evidence.
Python / Inert snippet
def sumar(a, b):
return a - bThe name suggests addition.
The written operator is subtraction.
Claim 01 / Structure
sumar is addition (Add).CONTRADICTED
The snippet contains subtraction (Sub). The structure contradicts the declared operator.
Claim 02 / Behavior
INSUFFICIENT_EVIDENCE
Inspecting this structure does not prove general behavior or replace execution tests.
Prepared synthetic results. The engine does not run on this site. The example comes from the experimental branch and the snippet is displayed as text. Read the technical explanation
Project status
The main release and experimental development have different scopes. The example above belongs to the experimental exploration.
| Release | Scope | Code and documentation |
|---|---|---|
Main main | Querying registered JSON aggregates, checking hashes and limited consistency, generating and retrieving reports through a CLI, Python API and MCP. | View main branch ↗ Read documentation ↗ |
Experimental experiment | Reading sources and checking specific static claims. The demonstration uses synthetic data and is not integrated into main. | View experiment branch ↗ Documented experimental demo ↗ |
Limits and next steps
pgopcore is a prototype. A static property does not prove functional correctness, and a generated report does not approve a task. When evidence is insufficient, that limit is part of the result.