Your First Knowledge Base
A guided walkthrough — create entries, search, connect to Claude, and see everything working together.
This walkthrough takes you from zero to a working knowledge base connected to Claude. It assumes you’ve already installed Pyrite (see Quick Start if not).
Create a knowledge base
mkdir my-brain && cd my-brain
pyrite init --name my-brain
This creates a kb.yaml file and an empty directory structure. Let’s add some content.
Add entries
Create a few entries to work with:
# A person
pyrite create --type person --title "Sarah Chen" \
--body "Engineering lead at Acme. Considering move to consulting." \
--tags "team,engineering"
# A decision
pyrite create --type decision --title "Switch to async standups" \
--body "Decided 2026-03-01. Reduces meeting load by 3hrs/week." \
--tags "process,meetings"
# A note with links
pyrite create --type note --title "Q1 Planning Notes" \
--body "Discussed headcount with [[sarah-chen]]. Key decision: [[switch-to-async-standups]]." \
--tags "planning,q1"
Each command creates a markdown file with YAML frontmatter. Look in the notes/, people/, and decisions/ directories to see what was generated.
Search
Now search across your entries:
# Keyword search
pyrite search "async standups"
# Find entries by tag
pyrite search "tag:engineering"
# Semantic search (if you installed pyrite[semantic])
pyrite search "team changes" --mode=semantic
Explore connections
Pyrite tracks [[wikilinks]] automatically:
# What links to Sarah?
pyrite backlinks sarah-chen
# Get the full entry
pyrite get sarah-chen
# See a timeline of recent changes
pyrite timeline --limit=10
Connect to Claude
Add Pyrite’s MCP server to your Claude configuration:
{
"mcpServers": {
"pyrite": {
"command": "pyrite",
"args": ["mcp"]
}
}
}
Restart Claude Desktop (or Claude Code), and you can now ask Claude things like:
- “What do I know about Sarah Chen?”
- “Find all decisions made this quarter”
- “Create a note summarizing our Q1 planning”
- “What entries are linked to async standups?”
Claude will use Pyrite’s MCP tools to search, read, and create entries in your knowledge base.
Start the web UI
For a visual interface:
pyrite serve
# Visit http://localhost:8088
The web UI gives you:
- A rich text editor with
[[wikilink]]autocomplete - An interactive knowledge graph showing connections
- Search with filters and facets
- Collections for organizing entries into views
What’s in a KB?
After these steps, your knowledge base directory looks like:
my-brain/
kb.yaml # KB configuration and type definitions
notes/
q1-planning-notes.md
people/
sarah-chen.md
decisions/
switch-to-async-standups.md
Each file is plain markdown you can edit in any text editor. The YAML frontmatter stores structured fields (type, tags, links). Git tracks all changes.
The SQLite index (.pyrite/index.db) is derived — delete it and rebuild with pyrite index build. Your files are always the source of truth.
Next steps
- How Pyrite Works — understand the architecture
- Custom Types — define your own entry types
- MCP Server — configure access tiers for AI agents
- Deploy — run Pyrite on a VPS for your team