The problem
You have years of accumulated notes, READMEs, internal wikis, code comments. Useful information lives in there. Finding it costs you 5-10 minutes per lookup. RAG turns those minutes into seconds.
Recommended setup
| Agent | Hermes Agent or Nanobot, with your vector store behind an MCP server |
| Hardware | Mac Mini M4 (24 GB+) for fast local indexing, or any mini PC if you index occasionally |
| LLM | Local Mistral 7B Q4 or Llama 3 8B Q4 via Ollama for the response generation |
How it works
Index your notes into a vector store (Postgres with pgvector, or Qdrant) with a local embedding model, then give the agent a search tool over it through an MCP server, which Hermes Agent and Nanobot both connect to. On query the agent retrieves the most relevant chunks and sends them to the model with your question, citing the source files. Until October 2026 this page said ZeroClaw does all of this on a notes folder by itself; its RAG module covers hardware datasheets only.
Reality check
No benchmark of ours here: indexing time and query latency depend on the size of your notes, the embedding model and the hardware, so time a first index on your own corpus. Answers are useful when the relevant content is in the corpus and useless when it isn't; the source citations let you check which. Re-indexing can be incremental, so day-to-day edits do not need a full rebuild.
What breaks
- Documents in obscure formats (proprietary databases, scanned PDFs without OCR)
- Queries that need synthesis across many chunks (RAG is good at retrieval, average at synthesis)
- Outdated content if you don't refresh the index
Alternative setups
Hermes Agent + Postgres with pgvector if you want the vector store separately for other tools to query. Nanobot with an MCP server for your document store if you want a lighter personal agent.
Agent defaults read in the projects' own documentation on 2026-10-18: Nanobot README (v0.3.5), Nanobot security settings (v0.3.5), Nanobot docs (v0.3.5): configure MCP tools, Hermes Agent docs: use MCP with Hermes.