Production · Pi 5 · FastMCP
ShaneBrain MCP Server
34 tools. Local AI. No cloud required.
42
Tools
12
Groups
17
Weaviate Collections
8100
Port
Tool Groups
Knowledge (2)
- search_knowledgeSemantic search across LegacyKnowledge RAG
- add_knowledgeStore new knowledge with category and source
Chat (3)
- search_conversationsSemantic search across conversation history
- log_conversationLog a message to the Conversation collection
- get_conversation_historyRetrieve conversation by session ID
RAG Chat (1)
- chatFull RAG pipeline: search knowledge + Ollama generate
Social (2)
- search_friendsSemantic search across FriendProfile
- get_top_friendsList friends sorted by interaction count
Vault (3)
- vault_searchSearch PersonalDoc (medical, legal, financial)
- vault_addStore a new vault document with category and tags
- vault_list_categoriesList all vault categories with document counts
Notes (3)
- daily_note_addAdd a journal entry, todo, or reminder
- daily_note_searchSemantic search across DailyNote
- daily_briefingGenerate today's briefing via Ollama
Drafts (2)
- draft_createAI-generate a draft (email, post, letter) with vault context
- draft_searchSearch saved drafts by topic or type
Security (3)
- security_log_searchSemantic search across SecurityLog
- security_log_recentRecent security events filterable by severity
- privacy_audit_searchSearch PrivacyAudit records
Weaviate Admin (2)
- rag_list_classesList all Weaviate collections with object counts
- rag_deleteDelete a Weaviate object by collection + UUID
Ollama (2)
- ollama_generatePrompt the local LLM with optional system prompt
- ollama_list_modelsList downloaded Ollama models with sizes
Planning (3)
- plan_listList markdown files in a planning subfolder
- plan_readRead a planning file's full content
- plan_writeCreate or append to a planning file
System (1)
- system_healthPing Weaviate, Ollama, Gateway — return latency dashboard
Stack
FastMCP
Streamable HTTP on port 8100. MCP JSON-RPC at /mcp, health at /health.
Weaviate 1.36.2
17 collections, nomic-embed-text 768-dim vectors, Docker on shanebrain-network.
Ollama 0.17.7
Local inference — shanebrain-3b, llama3.2:3b, 4-node cluster proxy on port 11435.
Raspberry Pi 5
16 GB RAM, 2x 2TB NVMe RAID 1, running everything local-first.
Pydantic v2
All 42 tool inputs validated. MCP annotations on every tool.
Claude Integration
Register with: claude mcp add --scope user shanebrain --transport http http://localhost:8100/mcp