01Where to start
One engine. Three ways to consume it.
Run it yourself for free, let us host it for your team, or have us deliver the finished study. The engine is the same — you choose the effort.
Open Core
For builders, PMs & researchers who live in an AI agent
Run the engine yourself — free, local, open.
ExploreCloud
For teams who want it hosted, shared & governed
Hosted, for teams — the same UI, managed for you.
ExploreResearch
For teams with no time to run it themselves
Done-for-you studies — buy the finished report.
Explore02Why it’s different
Debates. Disagrees. Remembers.
One population, many jobs.
Swap positioning, pricing, demand, ideation, discovery — without rebuilding the panel.
Method
- Positioning
- Pricing
- Demand/JTBD
- Ideation (HMW)
Grounded in evidence.
Every verdict carries the quote behind it — a stance from lived experience, not a vibe.
Marcus Hale
CFO · scale-up
“Per-seat pricing punishes exactly the teams you want expanding. I’d cap it.”
It remembers — and shifts.
Bi-temporal memory: who supported, who balked, and what changed their mind.
Sentiment across the council
03How it works
From persona to synthesis.
Persona
A longitudinal profile grounded in memory — not a one-shot prompt persona.
Simulation
Personas live out workdays and accumulate experience over time.
Council
A moderated debate — tension, pain-discovery, then a convergence loop.
Synthesis
One auditable answer, with evidence quotes and shift-tracking.
04Solutions
All solutionsPositioning
Does the message land — or bounce?
Pricing / Price-Sensitivity
Willingness-to-pay without leading the witness.
Demand / Jobs-to-be-Done
Why people switch — the forces of progress.
Ideation (How-Might-We)
From a How-Might-We to concepts worth building.
Continuous Discovery
A standing panel that remembers across weeks.
Churn Reasons
Why they leave — and which reasons you can fix.
A/B Test
Which variant wins — and who it wins for.
05The architecture
The generative-agents lineage — built to deliberate.
Memory, reflection, planning — the Stanford generative-agents architecture (Park et al.), where the 1,000-person study reached 85% accuracy. We’re that lineage, not the flat “synthetic users” one.
A bi-temporal memory graph, a moderated council, and a synthesis that you can audit — running locally, so your data never leaves the room.
Explore Open Core06Get started
Bring your own AI. Run it locally — free.
We don’t sell tokens. The free plan runs unlimited councils on your own Claude Code, Cursor or Codex — one line, no API key, set up in under a minute.
claude mcp add sonaloop --scope user --transport http …Opus 4.8 · runs locally · your data stays in the room
Point your agent at it
claude mcp add sonaloop --scope user --transport http https://api.sonaloop.com/mcpTrust / FAQ
Why is local free?
Your own AI agent writes the text. We sell methodology, collaboration and governance — not tokens.
Do I need an API key?
No. Only the optional extras — avatars and semantic recall — use one.
Does my data leave my machine?
Never on Open Core — personas, councils and syntheses all live on your machine.