Sitearc
Start free
Measure and learn

What each change taught you, kept for the next proposal.

Learnings run from measured to inferred, and the server sets the tier. Your agent reads them before it proposes anything.

Start freeWhat your agent can do
Demo data. Harbor & Pine is a fictional boatyard.

What it shows

  • 01

    Evidence tiers

    Four tiers: measured, measured on a narrower scope, observed, and inferred.

  • 02

    The source

    Each learning names the experiment behind it, or says there is none.

  • 03

    This site or every site

    A learning holds for one site, or for every site you run.

  • 04

    Inferred is marked

    An inferred learning is an untested hypothesis, and it says so.

  • 05

    Search and filter

    Search learnings, or show only the tiers you trust.

  • 06

    Shown in Work

    Related learnings appear beside each new proposal in the drawer.

What your agent can do

Your agent works with Learnings through tools like these. Some only read. Others may change something or spend credits.

Every tool in the MCP referenceHow agents connect
app.sitearc.ai/mcp4 of 130 tools
  • search_learningsReads

    Reads this site's learnings and the ones that hold for every site, tier first.

  • record_learningMay change or spend

    Records a learning. The server sets its tier, never the agent.

  • get_project_briefReads

    Starts a session with the site, its repo, its sources, its learnings and running experiments.

  • create_proposalMay change or spend

    Records a change with its evidence and a prediction. It ships nothing.

How it works

  1. 1

    An experiment concludes, and your agent records what it learned.

  2. 2

    The server sets the tier from the measurements. Repeating a claim never raises it.

  3. 3

    The project brief your agent starts a session with includes every active learning.

  4. 4

    Related learnings show beside each new proposal, so you see what worked before.

Limits and credits

Know the edges before you rely on it. Credits pay for outside data and heavy work, and saved data never costs any.

Limits

  • Your agent can't choose a tier. Only a concluded experiment earns tier 1 or 2.
  • An inconclusive experiment gives a tier 4 learning.
  • Superseded and retracted learnings are kept, but not shown.

Credits

0This feature spends no credits. Reading saved data is free on every plan.

One credit is $0.001. Pro includes 100,000 a month, and Free includes none. See pricing

Connect your first site free.

Add every site you run, at no cost. Your agent can start the same day.

Start freeSee pricing