Version-controlled OKF toolkit for .NET with Native AOT support
Store AI agent memory as a Git-tracked directory of Markdown and YAML, validate bundles with a CLI validator, and render browsable documentation. OKF4net is a Windows .NET toolkit by Julien CHABLE for developers and researchers building auditable, text-first agent memory. It supports Native AOT deployment, a built-in YAML-subset parser, MCP integration, and a static-site renderer, fitting .NET engineers who need transparent, file-based knowledge management for LLM-driven workflows.
What does the tool do for AI knowledge management?
OKF4net treats knowledge as a version-controlled folder of Markdown files with YAML front matter, converting that structure into machine-readable context for agents. It provides a CLI validator named okf to audit bundles and an okf-render component to produce a browsable static site from the same source files. The tool includes a Markdown link scanner and a YAML-subset parser, so the knowledge bundle can be validated and presented without external parsers.
How does it affect runtime and deployment on Windows?
The toolkit supports Native AOT, which produces standalone binaries that start quickly and use smaller runtime footprints compared with typical just-in-time builds. It targets current .NET framework versions and runs as a portable CLI or via WinGet. Because it is implemented with no external dependencies, the runtime surface is limited to the .NET Base Class Library, reducing extra assemblies at deployment time.
Is it safe to use and auditable in production workflows?
The tool enables "git-blameable" agent memory, giving a clear audit trail for each file change and its author. The CLI validator enforces OKF v0.2 conformance before serving bundles, which helps prevent malformed metadata from reaching an agent. Integration points include the Model Context Protocol and a Microsoft agent framework adapter, so the knowledge can be consumed by established agent pipelines while preserving file-level provenance.
Do developers need specialist knowledge to operate it?
The toolkit targets software engineers and .NET developers; it expects familiarity with Markdown, YAML front matter, and Git workflows. Command-line operation and Native AOT publishing require basic .NET SDK experience. In addition, the built-in parsers and renderers let teams use standard text editors for authoring, while the validator provides a clear checklist for repo hygiene without adding third-party libraries.
Good fit for .NET engineers who want auditable, file-based agent memory
The tool suits developers who prefer Git and plain-text source control for agent context, and who can work with CLI workflows and the .NET SDK. Expect a small deployment surface because the implementation avoids external packages; the main trade-off is that authors must maintain Markdown/YAML discipline and Git practices. As a practical tip, run the validator as part of your CI checks before publishing bundles to agents.
