01
Agent Registry
Declarative YAML contracts for AI agents: allowed actions, blocked actions, budgets and approvals. Ed25519-signed bundles, runtime enforcement, LangChain and CrewAI wrappers. On PyPI.
Open source by Hlinor
Tools that make AI work provable.
Every Hlinor audit runs on tools we publish and maintain in the open. Declare what your AI agents may do, verify what is actually running, and keep contracts aligned with reality - inspectable, deterministic, free to audit yourself.
Open source is our evidence: the method behind every audit report is public and reviewable.
The toolchain
Each tool works on its own. Together they cover the full path from declared intent to accepted evidence.
02
Agent Control Plane
Read-only scanner that inventories agents, models, tools and MCP servers in a codebase, then flags deterministic risk findings with file/line evidence. SARIF output for CI gates.
03
ScopeGuard
Finds scope drift between the signed agreement and the message thread before it eats your margin. Deterministic rules, evidence trail, human decisions. Hosted pilot available.
Why open source?
An audit you cannot inspect is an opinion. Our detection rules, policy formats and scanners are public, so your engineers can verify exactly how findings are produced - before trusting the report.
What this gives you
- No black box: the method is on GitHub, reviewable line by line.
- No lock-in: YAML policies and SARIF reports work with your existing tooling.
- Self-service first: run the tools yourself, call us when you want independent review.
From tools to proof
Want an independent review on top of the tools?
We run fixed-scope governance reviews of AI agents and workflows: policy inventory, runtime permission checks, audit-log gaps and a client-ready report in 72 hours.