Early-access pre-sale - founding price

Ship decisions backed by evidence.

Agent skills, A/B loops and tamper-evident work logs for people who'd rather prove than guess.

Runs on your machine. Python 3.9+, zero dependencies. Works with Cursor and Claude Code.

Two kits. One habit:
evidence first.

Private GitHub repo access, instantly after checkout. Pay once, keep every update in v0.x.

For solo developers
AK
Agent Kit™

Build decisions on evidence.

Your coding agent cites sources, runs A/B tests and asks for a calibrated go/no-go before you ship.

  • 4 Cursor & Claude skills: evidence-first, A/B loop, Jev decision, ship-small
  • MCP config: filesystem, git, fetch, memory
  • agesoma CLI: evidence log, A/B stats (p-value, P(B>A), sample size), revenue per visitor
  • Playbooks: weekly loop, first pricing test, customer calls
$49one-time · single user
Get the Agent Kit
For teams
PK
Proof Kit™

Your team's best work, on the record.

Everyone keeps a dated work log they own. The team gets its DNA - mix, gaps, single-owner risks - from what people choose to share.

  • Evidence-log agent: proof add, git import, weekly summary
  • Hash-chained log: edits are detectable, exports are verifiable
  • Team DNA report from voluntary exports - no ranking, ever
  • Rollout playbook & templates built around "the log is yours"
$99one-time · up to 10 seats
Get the Proof Kit

Payments handled by Polar (merchant of record). 7-day money-back guarantee: ask for a refund within 7 days of purchase, no questions asked (per Brazil's Consumer Defense Code, art. 49).

AGESOMA™

It takes evidence to win.

How it works

Three minutes from checkout to your first logged decision.

01

Get repo access

Checkout gives your GitHub account access to the private repo.

02

Install in one line

./install.sh cursor or proof init. Skills, rules and MCP land in your project.

03

Measure, then decide

Log evidence, run the A/B, let the winner become your next baseline.

$ agesoma ab read price
price metric=paid_rate control=49
        49: 6/160 = 3.75%  rev/visitor=1.84
        99: 5/158 = 3.16%  p=0.775  P(beats control)=39.71%  rev/visitor=3.13
verdict: no clear winner yet - keep running or cut as a tie

Questions

Is this a SaaS? Where does my data go?

No servers. Both kits run on your machine and store plain JSON files. Nothing is sent anywhere unless you call an API yourself.

Isn't the Proof Kit employee surveillance?

It's built to be the opposite. The log lives on the employee's machine, private entries never export, and the team report refuses to rank people. Adoption is voluntary by design.

Do I need Jev or a paid model?

No. Everything works offline. The Jev decision step and LLM drafting are optional and work with free models through any OpenAI-compatible gateway.

What's the license?

Agent Kit: one person, unlimited own projects. Proof Kit: one organisation, up to 10 seats. You can modify it; you can't resell it.