okft
Provides tools for AI agents to navigate and query OKF knowledge bundles through overview, read, search, and list operations.
README
okft
Lint and serve Open Knowledge Format (OKF) bundles.
OKF is Google's open spec for representing organizational knowledge as a
directory of markdown files with YAML frontmatter — a knowledge graph that
both humans and AI agents can read natively. okft covers the two
sides of keeping a bundle healthy and useful:
okft lint— validate a bundle against the OKF v0.1 spec, plus hygiene checks (broken links, orphaned concepts, malformed timestamps). Wire it into CI so your knowledge bundle can't rot silently.okft serve— expose a bundle to any MCP-capable AI agent (Claude, Gemini CLI, Cursor, …) as a set of navigation tools: overview, read, search, list. Deterministic graph traversal, no embeddings, no database.
Install
pip install okft # lint only
pip install 'okft[serve]' # lint + MCP server
Lint
okft lint path/to/bundle
analytics/tables/orders.md:1 error E003 frontmatter must include a non-empty `type` field
analytics/metrics/churn.md:24 warning W001 link target does not resolve in bundle: /analytics/tables/order
engineering/runbook.md:1 warning W004 concept is never linked from any other document
12 concepts checked: 1 error(s), 2 warning(s)
Exit code is 1 on errors (or on warnings with --strict), so it drops
straight into CI. --format json emits machine-readable findings.
Rules
| Code | Severity | Check |
|---|---|---|
| E001 | error | concept file has no YAML frontmatter block |
| E002 | error | frontmatter is not parseable YAML |
| E003 | error | missing or empty type field |
| E004 | error | reserved file (index.md / log.md) has frontmatter |
| W001 | warning | link target does not resolve inside the bundle¹ |
| W002 | warning | timestamp is not ISO 8601 |
| W003 | warning | tags is not a list of strings |
| W004 | warning | orphan concept — nothing links to it (--no-orphans to skip) |
| W005 | warning | bundle root has no index.md |
| W006 | warning | log.md headings are not ISO 8601 dates |
| W007 | warning | concept has no title |
¹ The spec requires consumers to tolerate broken links, so they are warnings, never conformance errors.
Both standard markdown links (/analytics/tables/customers.md, relative
paths) and [[wiki-style]] links are resolved.
Serve to an AI agent
okft serve path/to/bundle
Runs an MCP server (stdio) with four tools:
| Tool | Purpose |
|---|---|
okf_overview |
root index + every concept grouped by type |
okf_read |
one concept: frontmatter, body, outbound & inbound links |
okf_search |
ranked full-text search with snippets |
okf_list |
filter concepts by type and/or tag |
Register it with Claude Code:
claude mcp add acme-brain -- okft serve /path/to/bundle
or in any MCP client config:
{
"mcpServers": {
"acme-brain": { "command": "okft", "args": ["serve", "/path/to/bundle"] }
}
}
Try it
A small example bundle ships in examples/acme_brain:
okft lint examples/acme_brain
okft serve examples/acme_brain
CI example (GitHub Actions)
- uses: actions/setup-python@v5
with: { python-version: "3.12" }
- run: pip install okft
- run: okft lint knowledge/ --strict
Status
Tracks OKF v0.1. The spec is young and so is this tool — issues and PRs welcome, especially reports of real-world bundles that lint incorrectly.
License
Apache-2.0
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.