StackMap
Subscribe
Explore / arkon
nduckmink

arkon

Self-hosted enterprise knowledge hub + MCP server: an LLM pipeline compiles SOPs and docs into a traceable, human-reviewed wiki, then serves it to AI clients scoped by department and role.

1,187 256 Python NOASSERTIONupdated 2 months ago
View on GitHubDispute this mapping →
Curator's take

The MRP pipeline is the differentiator: instead of chunk-and-index, documents are compiled into merged, source-traceable wiki pages behind a plan-review gate — knowledge management, not just retrieval — with department-scoped serving over one MCP endpoint. When NOT: the license is NOASSERTION, a real blocker for the enterprises it targets — clarify before adoption; ★1.2k young, and the compile-everything approach costs LLM tokens where a plain vector store costs none.

Mapped by ShipWithAI editors · links verified
README.md

Arkon - The Open-Source Enterprise AI Knowledge Hub & MCP Server

License GitHub Stars Docker

Arkon

Arkon is a self-hosted, enterprise-grade knowledge management layer that bridges organizational data and AI clients. It runs as a centralized MCP Server (Model Context Protocol), compiling your SOPs, policies, and internal docs into a structured, traceable knowledge wiki - then serving that wiki to Claude and other LLMs through a single permission-scoped endpoint.

🚀 Why Arkon?

In most organizations, AI adoption is fragmented. Employees copy-paste documents into chatbots, producing inconsistent context, security risks, and duplicated work.

Arkon treats AI as a managed organizational resource. Every employee gets the right context, automatically and securely - filtered by their department and role.

How Arkon Works


✨ Key Features

🧠 Intelligent Knowledge Wiki - the MRP Pipeline

Unlike a vector database that just chunks and indexes, Arkon's MRP pipeline (Map → Reduce → Plan-review → Refine → Verify → Commit) actually compiles documents into a coherent wiki of interlinked pages.

  • Plan review before write: every ingestion produces a human-reviewable plan listing which wiki pages will be created or updated. Editors can regenerate the plan with feedback before any page is written.
  • Page merge instead of overwrite: when a new source touches an existing wiki page, content is LLM-merged so prior knowledge is never lost.
  • Traceable claims: every page records the source documents it was compiled from.
  • Image-aware: vision captions are baked into source text before compilation, so wiki pages reference the right images in the right places.
  • Resumable: drafts persist mid-pipeline; a crashed run resumes without re-doing the expensive LLM work.

📚 Wiki Browser & Knowledge Graph

  • Three-panel layout: page tree, content, backlinks & outlinks.
  • Full-text + semantic (pgvector) search.
  • Interactive knowledge graph visualization (per-scope or global).
  • Wikilink-style cross-references between pages.
  • Version history and rollback on every page.
  • Draft proposal → editor review → approval workflow.

Arkon Wiki Browser UI — three-panel knowledge base with page tree, content viewer, and knowledge graph

🏢 Department & Global Scopes

Precise context boundaries with their own scoped wiki, document set, and department rosters.

  • Department-level isolation for HR, Legal, Engineering, etc.
  • Global scope for company-wide SOPs, policies, and general knowledge.
  • Hard scope enforcement: members only see knowledge from their assigned departments or the global realm - at the API, MCP, and search layers.

🛂 Fine-Grained RBAC

Role-based access control at the department and global levels.

  • Built-in roles: Viewer · Contributor · Editor · Admin to control draft, edit, and setting access.
  • Granular permissions (doc:read:own_dept, wiki:edit:all, org:settings:manage, ...).
  • Audit log for every privileged action - settings changes, plan approvals, and role updates.

🔌 MCP Server for Claude & Other AI Clients

Employees connect Claude Desktop or Claude.ai to Arkon via OAuth 2.1 + PKCE — just add the server URL and sign in through the browser. No manual token copying required. The MCP server exposes:

  • **Wiki too

Continue your stack

What teams reach for next — and why each earns a place beside arkon. Ranked by curator confidence.