Start by saving a password. End up with a personal AI that knows your life, runs on your machine, and answers to no one but you.
↓ Pick your starting point
No Ollama. No API key. No subscription. Tier 0 gives you an encrypted vault, a private credentials manager, memory capture, a scheduler, and full Telegram access — all completely offline.
Store usernames, passwords, URLs, and notes — all encrypted in your vault behind a passphrase only you know. Import your existing passwords from LastPass, Bitwarden, or any browser export (CSV). Export back to encrypted .pfcreds format or CSV. The people who built Pathfinder have never seen your passwords. And never will.
password: netflix and your bot privately returns the password — no app to open, works from anywhere.
Connect Pathfinder to a private Telegram bot you own. Once set up, you can save memories, journal, search your notes, retrieve passwords, and capture photos — all from your phone, anywhere in the world. No public server. No third-party app. Just a private Telegram bot talking to your own machine.
Save anything to your encrypted local memory. From the desktop or from Telegram:
memory: called the dentist, appointment April 20 or
journal: great meeting today with the team.
Everything is tagged by date, stored locally, and searchable instantly.
Your life archive, owned by you.
Create recurring tasks that run on a schedule and deliver to Telegram or memory. At Tier 0, use the scheduler for non-AI reminders and journal prompts. At Tier 1 and above, the scheduler calls your local AI model and sends morning briefs, news digests, and research summaries — automatically, while you sleep.
Install Ollama (free), pull a model, and Pathfinder starts thinking. Conversations, tool use, scheduled briefs — all running privately on your hardware. No API account. No per-message cost. No data leaving your machine.
Ask questions, get answers. Your conversations never touch a cloud server — the AI runs on your machine using your GPU or CPU. Works with no internet connection. Recommended models: qwen3 for reasoning and tool use, gemma4:e4b for best quality on an 8GB GPU, llama3.2 for fast general chat.
BitNet b1.58 is a 2-billion parameter model that runs entirely on your CPU using Microsoft's custom 1.58-bit inference engine. No graphics card needed. Uses only 0.4GB of memory. If your machine doesn't have a dedicated GPU, BitNet is your local AI option. Requires a one-time build from source — ask Trailhead "how do I install BitNet?" for step-by-step instructions.
ollama pull downloads correctly but returns empty responses — Ollama does not support BitNet's custom kernels. Use the bitnet.cpp install for working inference. Ask Trailhead for the full guide.
Drop any Markdown file into your docs\ folder and Pathfinder
indexes it automatically on startup. Ask the AI "what did I write about my
insurance policy?" and get an answer from your own files — no sending
documents anywhere. Trailhead uses this same system to answer questions
about Pathfinder itself.
Every morning at 7am (your local timezone), Pathfinder runs your local AI model, asks it to summarize the overnight news, and sends the result to your Telegram. No cloud involvement. No API cost. Your AI works while you sleep. The Morning Market Brief and Iran War Latest jobs shown below were running this way — locally, privately, on schedule.
Create separate AI identities for different parts of your life. Me-Personal uses local qwen3, casual tone, accesses your personal memories. Me-Work uses a formal system prompt and sees only work-related context. Me-Health uses a medical persona with access to your health journal. Switch between them via Telegram PIN — one message, different mind.
Skills are folder-based workflow pipelines that run a sequence of AI steps and deliver results to Telegram or memory. No coding required to use them. Included skills: Morning Brief (news summary), Research (web search synthesis). Edit the prompt files in the skill folder to customize for your needs.
Your vault stays local. Your memories stay local. You just borrow cloud intelligence when it's worth it — and Pathfinder tells you exactly what each call costs. No assumed provider. No hidden default. You choose what runs.
Pathfinder never assumes a provider. The model picker shows only what you have actually set up — installed Ollama models and providers with API keys. Nothing phantom, nothing pre-selected. Add a Google API key and Gemini appears. Add an Anthropic key and Claude appears. Remove the key and it disappears. Your configuration, your choice.
Every cloud AI call is logged in the Logs panel with an estimated cost in USD, calculated from provider rates and token counts. Local model calls show inference time in milliseconds and "Free." You always know what your AI is costing you — without logging into a vendor dashboard. New models on known providers inherit rates automatically — no configuration needed.
Add a Serper API key and Pathfinder can search the web and synthesize results through your chosen AI model. Used by scheduled jobs: the Morning Market Brief pulls live market data every morning before the open. The agent loop runs up to 5 search-and-reason iterations automatically.
Connect Pathfinder to Google Calendar, Gmail, and 400+ other services via n8n — a visual workflow builder that runs on your machine. Ask "what's on my calendar tomorrow?" and get an answer drawn from your actual calendar, processed locally. Pathfinder sends a structured payload to n8n; n8n handles the integration; results come back and the AI synthesizes them.
You can start at any tier and move up when ready. Nothing locks you in.
Install Pathfinder, create your vault passphrase, go to Settings → Credentials. Import your existing CSV export from any password manager. Done.
Download Pathfinder →Install Ollama from ollama.com. Run ollama pull qwen3 in PowerShell. Open Pathfinder, set persona to Scout tier, select qwen3. That's it — private AI, no cost, no account.
Install Pathfinder. Go to Settings → API Keys, add your Google or Anthropic key. Set persona to Pioneer tier. The model picker will show your configured providers. Start with Gemini — it has a free tier.
Download Pathfinder →