Skip to content

What is PieKBS

PieKBS is a local-first knowledge search engine for agents. It distills raw documents into a structured, reviewable Markdown wiki, then exposes three MCP tools — kb_search, kb_page, and kb_add — that let agents search, read, and write knowledge at their own pace.

Design Philosophy

PieKBS is built around one observation: agents use external knowledge tools the same way humans use search engines — they issue multiple queries from different angles, follow links, and synthesize their own conclusions. They do not want a pre-packaged answer; they want the raw materials to form their own.

This means PieKBS's job is not to answer questions. It is to make sure that when an agent searches for something, it finds the right documents — and can read them in full.

PieKBS vs RAG

RAGPieKBS
Knowledge formImplicit (vectors or chunks)Explicit (Markdown, auditable)
Agent rolePassive receiver of contextActive searcher and reader
Answer sourceSystem-generatedAgent-synthesized
AuditableNoYes — git diff, lint, conflict links
Multi-hop reasoningLLM-dependentGraph expansion via related links
EmbeddingRequiredNot required (pure FTS)

KB Structure

text
piekbs-kb/
  raw/              Source of truth — drop files here, watcher auto-distills.
  wiki/
    source-notes/   One distilled note per raw document. FTS search target.
    concepts/       Cross-document synthesis: concepts and methodologies.
    comparisons/    Cross-document synthesis: side-by-side comparisons.
    decisions/      Cross-document synthesis: technical decisions.
    _draft/         Pages with < 2 sources (not indexed yet).
  schema/           KB-local authoring rules and page templates.
  index/            Generated artifacts (SQLite FTS index). Do not edit.

Ecosystem

Learn More

Released under the MIT License.