Updated July 28, 2026
Getting started
Create a workspace, connect your first source, and run the first analysis — what happens at each step, and what starts (and doesn't start) your trial.
Read the pageSightspool Docs
Operate the system.
Short, current pages on running Sightspool — from the first connected source to the register, interventions and coding-agent access. The SDK’s API reference lives at sdk.sightspool.com.
Start here
Updated July 28, 2026
Connecting sources
What Sightspool reads, how your credentials are stored, and why sample data can explore the product but never back a published finding.
Read the pageThe register
Updated July 28, 2026
The assumption register
How a belief becomes falsifiable — question, signal, threshold, verdict — and how ruled verdicts write back into your coding agent's context file.
Read the pageCapabilities
Updated July 28, 2026
Demand probes
Test demand for a feature that doesn't exist yet — a dashed-border button where the feature would live, honest by construction, counted as votes in your register.
Read the pageUpdated July 28, 2026
Coding-agent access (MCP)
Connect Claude Code, Cursor or any MCP client to your workspace so your coding agent can check what's known — and deposit the assumptions it makes while building.
Read the pageUpdated July 28, 2026
Deep-capture studies
Bounded, cohort-scoped session recording in your own PostHog — justified by one assumption, approved by you, armed for a set window, then switched off again.
Read the pageGuarantees
Updated July 28, 2026
Honesty guarantees
The claims Sightspool structurally cannot make: unsourced numbers, verdicts below sample, agent-declared certainty, actions you didn't approve, or an undisclosed painted door.
Read the page