Bridging the Gap Between AI Agents and Enterprise Context
Stop feeding raw database dumps or arbitrary vector chunks to LLMs. Represent your metadata, tables, schemas, and APIs as a connected, type-safe Markdown knowledge graph with OKF.
High-Performance CLI Utilities
Experience the OKF-go command-line interface. Switch tabs below to see simulated runs of our core subcommands.
Context Assembly, Visualized
Vector databases fetch fragmented snippets out of order. OKF works differently: starting from a concept, it performs a Breadth-First Search (BFS) along explicit relative Markdown links.
It pulls in adjacent nodes (like database schemas referenced by tables, or APIs writing to columns) until the target character budget is filled.
tables/orders...tables/users (found in body).playbooks/database_cleanup.Live Concept Parser Playground
Edit the OKF document YAML frontmatter or body below. Watch our conformance compiler parse, check links, and display structured concept tokens in real-time.
# Transactions Table This stores historical transaction logs for payouts. It references the [Users Table](tables/users.md) through the `user_id` field. ## Key Fields * `id` (UUID): Primary key. * `user_id` (UUID): Failsafe foreign key reference. * `amount` (DECIMAL): Transaction value.
Built for Modern AI Integration
Every element of OKF is optimized to bridge standard development cycles with autonomous agent workflows.
Conformance Engine & Linter
Run okf lint to validate YAML schemas, catch unparseable headers, and trace all internal links to prevent disconnected or dead paths.
Metadata Harvesters
Auto-extract schemas. Point okf harvest at PostgreSQL, MySQL, Spanner, BigQuery, OpenAPI, Git repos, or Web pages.
Dynamic Context Assembly
Compile subgraphs instantly. Use BFS link traversal to query surrounding information and output XML blocks structured for direct LLM ingestion.
Model Context Protocol (MCP)
Serve knowledge directly. okf mcp supports Stdio/SSE modes to expose resources, tools, and prompts directly into Cursor, Claude Desktop, and Antigravity.
Interactive HTML Compiler
Build static developer docs. Run okf doc to compile your Markdown knowledge bundle into a fully searchable static website with relationship graphs.
IDE Language Server (LSP)
Get real-time validation. okf lsp implements a Language Server over Stdio, flagging missing metadata fields and broken links inside VS Code or Cursor as you type.
Bi-directional Sync Daemon
Unify knowledge bases. okf sync synchronizes concepts dynamically with external sources like Google Drive, Notion, Confluence, and Jira.
Bundle Operations
Compare and merge. Use okf diff and okf merge to compare bundles for differences or unify distributed knowledge bases.
AI Enrich
AI-powered organization. Use the --ai-enrich flag during harvesting to automatically generate descriptions, tag concepts, and enrich metadata using Gemini and any other models.
Knowledge Export
Export your knowledge graph in multiple formats. Run okf export to compile your entire OKF knowledge graph into standard schema.org JSON-LD format for easy sharing.
Getting Started
Get up and running with the OKF-go tool in seconds.
1. Quick Installation
Run the direct shell script installer to fetch the precompiled binary for macOS or Linux:
curl -sSfL https://okfgo.dev/install.sh | shThe script auto-detects architecture (amd64, arm64) and locations (e.g. /usr/local/bin or ~/.local/bin).
2. Build From Source
If you have Go installed (v1.26 or higher), you can compile the binary locally:
git clone https://github.com/abcubed3/okf.git
cd okf
go build -o okf main.go
./okf help3. Connect to Claude Desktop
To use the OKF bundle directly inside Claude Desktop as an MCP resource, update your configuration:
{
"mcpServers": {
"okf-knowledge": {
"command": "/usr/local/bin/okf",
"args": ["server", "--bundle", "/path/to/my-bundle"]
}
}
}