Find the right person
Ask who in your network knows about fundraising or Kubernetes and get ranked people with the evidence.
WhatsApp context CLI
Your network lives in your chats. wactx turns it into a private graph of who you know and what they know.
wactx syncs your WhatsApp history into a single DuckDB file on your machine, builds a graph of people, groups, topics and entities, and answers questions like “who in my network knows about fundraising?” with ranked people and the messages behind them. Your agents can query it too.
- Language
- Python and Go
- Storage
- DuckDB with vector and graph extensions
- Licence
- MIT
291Kmessages indexed in our own graph
15Kpeople in the relationship graph
105daily digests written from it since April
0.5 sfor a five-variant vector search
Why we built it
Introductions, decisions and commitments happen in chat threads, then scroll away. A founder with hundreds of groups has no way to ask who knows what.
We built wactx when we needed the right go-to-market advisors from tens of thousands of our own messages and no tool could answer the question. Now it writes our digest every morning and gives our agents the context they need.
What it does
- One QR scan to sync every chat, then incremental and live sync; it only reads, and never sends
- Everything in one DuckDB file you own, removed with a single command
- Hybrid search that fuses keyword and vector results, rewrites your question into variants, and embeds each message with its conversational neighbours
- A relationship graph of people, groups, topics and entities, ranked with personalised PageRank and community detection
- Local entity extraction for people, organisations, technologies and events, at no API cost
- Keyword, chat and date filters, with fast, balanced and deep search modes
- JSON output and a bundled Claude Code skill for agents, with any OpenAI-compatible model, including a local Ollama for fully private runs
Where it’s going
- One private graph across WhatsApp, Gmail, Telegram and Slack, with people matched across sources
- A native MCP server, so any agent can reach your context
- Search that tunes itself against evaluation metrics
- Relationships that carry time, so the graph shows how your network changes
Who it’s for
Daily digests
An agent summarises replies you owe, live work threads and what your communities are saying, every morning.
A CRM with no data entry
Build a relationship map and warm-intro paths from conversations that already happened.
Get started
uv tool install whatsapp-ctx-cli # or: pip install whatsapp-ctx-cli
wactx init # choose a provider and API key
wactx sync # scan the QR code on first run
wactx index # embed messages
wactx search "who knows about fundraising"Full documentation is in the repository README.
Work with us
Tell us what the work is, what data it touches and what constraints apply. We’ll work out what to build and how to measure it.