Resources · LLM Wiki CRM

Your CRM is a folder of markdown files. An LLM maintains it for you.

Drop meeting notes in a folder. The LLM extracts contacts, companies, deals and action items — updates every page, maintains cross-references, keeps your pipeline current. Zero lock-in. Your data is just files.

Built on Andrej Karpathy's LLM Wiki pattern →

  • No forms. No fields. No data entry.
  • Works with Claude, Cursor, or any LLM that reads an instruction file.
  • Git-versioned. Your history is auditable by default.
  • Self-hosted on your laptop or a private repo.

What you get

  • Complete folder scaffold — contacts, companies, deals, interactions, tasks.
  • A working CLAUDE.md instruction set that drives ingest, query, update and lint flows.
  • Dataview queries for pipeline, activity, stale-deal detection.
  • Templater presets for every entity type.

How it works — the four workflows.

I.

Ingest

Drop a raw meeting note in _inbox/. The LLM extracts contacts, companies, deals, interactions and tasks — creating and linking files automatically.

II.

Query

Ask plain-English questions. “What's our pipeline value?” “Warmest leads?” The LLM synthesises answers with citations across every file.

III.

Update

“Move the Acme deal to negotiation.” “Mark Jane inactive.” The LLM finds the file, updates YAML frontmatter, logs the change.

IV.

Lint

Request a health check — stale deals, missing data, broken links, overdue tasks. Structured report, suggested fixes.

Want team deployment?

Multi-user, audit trails and a shared Git remote. We'll set it up end-to-end.

Talk to us

The CRM your team will actually use.

Free and open source. Works with any LLM agent. Ships today.