How it works
One people layer.
Agents and humans both consume it.
Founders shipping AI dev tools who are already posting demos on X and have a GitHub repo with 500+ stars
Try a template
Step 1
Query in natural language or by API
An agent calls the API, or a recruiter types in plain English. Same engine, same result — no boolean strings, no rigid filters.
Step 2
Resolved across the open web
GitHub, X, LinkedIn, Reddit, Discord, personal sites. Sonner pulls fresh data with source attribution — not stale CSV exports.
Writing about distributed state. Author of tokio-state.
Recent activity
shipped tokio-state v3 today — 40% fewer allocations on consistent reads. writeup soon 🧵
SREcon talk recording is live — "State at Scale: when CAP stops being your friend"
Step 3
Structured profiles, not strings
Each person comes back as a typed object: identity, recent work, network, signals, inferred intent, freshness. Reason over it, or render it.
Hey Maya — saw the tokio-state v3 post. The allocation numbers are wild. We're wrestling with a similar state-layer problem at Sonner and would love to compare notes over a call
Step 4
Outreach an agent or a human can send
Personalized drafts that reference the person’s actual work. Dispatched programmatically through the API or sent from the workspace.
Built for the
agents deciding.
Structured people context for AI agents — and the humans working alongside them.