claudia

A thinking partner who tracks relationships, not just tasks.

Remembers your people. Catches your commitments. Learns how you work.

Install · Why Claudia · Her Mind · Integrations · How It Works

The Problem With AI Assistants

You tell ChatGPT about Sarah on Monday. By Wednesday, it's forgotten.

You make a promise in a meeting. Nobody tracks it. You promise a deliverable on Friday and lose track of it by Wednesday. You meet someone important, and three weeks later you can't remember what you talked about.

AI tools don't have memory. Claudia does.

What Makes Claudia Different

How Her Mind Works

Quick Start

1. Open the folder you want Claudia to live in with Codex, Claude Code, or Grok.

2. Paste this into the agent:

Install or upgrade Claudia in this folder. Run npx get-claudia and keep going until setup and health checks pass.

The installer detects whether it was launched by Codex, Claude Code, or Grok and targets the folder already open in that agent. In an empty or existing project folder, it installs Claudia there. In an existing Claudia workspace, it upgrades the framework and host integration in place while preserving your context, people, projects, and memory.

npx get-claudia also checks npm's latest release before changing files. If npx supplied a stale cached copy, Claudia immediately re-runs the current release. You do not need to remember @latest.

Prefer to run the command yourself? It is the same one-liner:

npx get-claudia

3. Start a new chat in the same folder. On a new Codex install, Claudia's initial onboarding is required and runs before substantive work. After onboarding—or after an upgrade with an existing context/me.md—she resumes normally. Inside the folder, the agent identifies itself as Claudia regardless of the underlying runtime.

That's the whole required setup. The plugin's AGENTS.md startup contract and memory tools load automatically in the new chat.

Optional: run /hooks once if you want Claudia's automatic start-of-session briefing and end-of-session transcript capture. Claudia still runs without trusting those hooks; you can ask for a briefing or save a memory directly.

npx get-claudia
cd claudia
claude

What's next:

• /morning-brief to see what needs attention
• Tell her about a person and she'll create a relationship file
• Share meeting notes and she'll extract action items
• npx get-claudia google to connect Gmail, Calendar, Drive, and more

Requirements: Codex or Claude Code, Node.js 18+, Python 3.10+ (for memory), Ollama (for embeddings)

Embeddings model: After installing Ollama, pull the required model: ollama pull all-minilm:l6-v2

npx get-claudia my-project --no-memory

cd your-claudia-directory
npx get-claudia

See It in Action

You:     What do I know about Sarah Chen?

Claudia: Sarah Chen is CEO at Meridian Ventures. You met at SaaStr 2025.

         📊 Relationship Health: Strong (last contact 3 days ago)

         🎯 Open Commitments:
         • Send updated pitch deck to Sarah [OVERDUE - was due Friday]

         💡 Recent Context:
         • She's interested in AI infrastructure angle
         • Wants to see technical architecture doc
         • Bringing in technical partner for next meeting

         🔗 Connected to: Series A (project), Priya Sharma (introduced by)

         Source: Call notes from Jan 28, Email thread Jan 30

Morning brief pulls everything together:

You:     /morning-brief

Claudia: ☀️ Morning Brief — Wednesday, Feb 19

         ⚠️ Needs Attention:
         • Send updated pitch deck to Sarah Chen [OVERDUE]
         • Quarterly review with Acme Corp is tomorrow

         👀 Watch:
         • Haven't spoken to Marcus in 18 days (usually weekly)
         • Two new commitments this week, zero closed

         📊 Today: 3 meetings, 2 open commitments

Judgment rules shape everything she does:

You:     I've got the Acme proposal and Marcus wants a catch-up.
         Which first?

Claudia: Acme proposal. Your judgment rules say revenue-generating
         work beats relationship maintenance when there's a
         deadline, and the Acme proposal is due Thursday.

         I'll remind you about Marcus tomorrow morning.

Pattern detection works across weeks, not just moments:

Claudia: I've noticed something across the last few weeks.
         You've taken on 4 new commitments without closing any.
         Last time this happened (early January), two slipped
         past deadline.

         Want me to flag the lowest-priority ones so you can
         decide what to defer?

Every conversation builds on the last. She remembers.

Who It's For

Claudia detects your work style and generates structure that fits:

Key Commands

Brain Visualizer

Launch with /brain to see your memory as a 3D network graph. Entities are nodes, relationships are edges, and everything is interactive: click to inspect, filter by type, search by name.

Integrations

Claudia works fully on her own, but integrations let her see further.

Google Workspace

Connect Gmail, Calendar, Drive, Docs, Sheets, Tasks, and more with a single setup command:

npx get-claudia google

This generates a one-click URL to enable all required Google APIs and walks you through OAuth setup. Three tiers available:

Obsidian Vault

Memory projects to an Obsidian vault at ~/.claudia/vault/. New installs default to the wiki layout: synthesized topic pages at ~/.claudia/vault/Wiki/, one per active person, project, or organization. Each page is written by Claudia from your raw memories, cites every load-bearing claim with [mem:NNN], and flags contradictions at the top. Obsidian's graph view connects them via [[wikilinks]].

Existing installs from v1.59 and earlier keep their PARA-organized vault (Active/, Relationships/, Reference/, Archive/) untouched. SQLite remains the source of truth; the vault is a projection you can browse, search, and read.

How It Works

45 skills · 33 MCP tools · 500+ tests

Claudia has two layers:

Template layer (markdown) defines who she is. 45 skills, rules, and identity files that Claude reads on startup. Skills range from proactive behaviors (commitment detection, pattern recognition, judgment awareness) to user-invocable workflows (/morning-brief, /research, /meditate, /wiki, /auto-research). Workspace templates let you spin up new projects with /new-workspace [name]. A built-in skill-router skill helps you discover what's available and disambiguates when a request straddles two skills.

Memory system (Python) defines what she remembers. Two daemon modes share the same SQLite database:

You ──► Codex / Claude Code ──► Reads Claudia's runtime contract ──► Becomes Claudia
                                                           │
                                                    MCP daemon (stdio)
                                                           │
                                                           ▼
                                                     SQLite + vectors
                                                           ▲
                                                           │
                                              Standalone daemon (24/7)
                                               ┌──────┼──────┐
                                               ▼      ▼      ▼
                                          Scheduler Ollama  Obsidian vault
                                                           (PARA structure)

Agent team for speed. Claudia delegates structured work to a two-tier team. Tier 1 (Haiku): fast workers for document archiving, processing, and schedule analysis. Tier 2 (Sonnet): a research scout with independent context for multi-turn web research. Claudia keeps relationship judgment and strategy decisions for herself.

Privacy and Safety

• Fully local. Memory, embeddings, cognitive tools run on your machine. No external APIs for storage.
• No external actions without approval. Every email, calendar event, external action requires your explicit "yes."
• Your data in two formats. SQLite database (~/.claudia/memory/) for fast semantic search, plus an Obsidian vault for reading and graph navigation. Two independent copies you own forever.
• Delete anything, anytime. Full control over your data. No lock-in, no cloud dependency.

System Requirements

Without the memory system, Claudia still works using markdown files. With it, she gains semantic search, pattern detection, and relationship tracking.

Ollama model: Run ollama pull all-minilm:l6-v2 after installing Ollama. This is the embedding model used for semantic search.

Platforms: macOS, Linux, Windows

curl http://localhost:3848/status
launchctl list | grep claudia
tail -20 ~/.claudia/daemon-stderr.log

launchctl load ~/Library/LaunchAgents/com.claudia.memory.plist

open -a Ollama          # macOS
ollama serve            # Linux

ollama pull all-minilm:l6-v2    # Embeddings (required)

rm ~/.workspace-mcp/token.json
# Restart Claude Code

cd your-claudia-directory
npx get-claudia .

Contributing

Claudia is source-available under the PolyForm Noncommercial License 1.0.0.

• Template (skills, rules, identity): template-v2/
• Memory daemon (Python): memory-daemon/ (tests: cd memory-daemon && pytest tests/)
• Installer: bin/index.js
• Architecture: ARCHITECTURE.md
• Dev guide: CLAUDE.md

License

PolyForm Noncommercial 1.0.0

Free for personal, research, educational, and nonprofit use. Commercial licensing: mail@kbanc.com

Created by Kamil Banc · @kamilbanc · AI Adopters Club

"I learn how you work. Let's get started."

If Claudia helps you, a ⭐ on GitHub means a lot.