SIMPSEN

Self-IMProving Software ENgine

Define your product vision. SIMPSEN's CEO agent autonomously plans features, orchestrates dev agents to build them in parallel, merges code, tests the deployed app, and loops—shipping tested code without human intervention.

Get Started

01

What SIMPSEN does

You define a product roadmap, mission, and CEO personality. SIMPSEN's CEO agent plans batches of 2–5 features, spins up parallel dev agents in isolated sandboxes, and ships code through quality gates—end to end.

After each batch, browser testing agents exercise the live app across multiple user personas. The CEO reviews test results and merge outcomes, then plans the next batch. Every decision is tracked in structured batch history that follows plan → build → merge → test → review.

From zero to shipped product, autonomously.

02

The intelligence loop

1
Ingest
2
Evaluate
3
Execute
4
Validate
5
Learn

continuous — every batch informs the next

Ingest — the CEO agent reviews your roadmap, mission, and past batch results to decide what to build next. Trigger events can also arrive via webhooks or the testing bot.

Evaluate — Claude Opus analyzes each trigger and generates a structured PRD with dependency-ordered user stories and acceptance criteria.

Execute — a sandboxed agent works through each story with circuit breakers—up to 3 attempts per story, 10 iterations per run.

Validate — quality gates enforce that typecheck and tests pass before any story is marked complete.

Learn — test results feed back to the CEO agent, informing smarter plans for the next batch.

03

Product intelligence

SIMPSEN tracks every batch from plan through ship, building a structured history of how your product evolves. Every change adds traceability:

Trigger
Plan
Build
Test
Ship

Tracked across every batch—CEO rationale, dev jobs, merge results, browser tests—the history surfaces patterns that no individual change could reveal.

Three views: Batch History for exploring past decisions, Job Logs to trace each agent's work, and Test Feedback to see what the browser testing agent found.

04

Trigger sources

The CEO agent and testing bot drive the core loop today. The webhook endpoint supports additional sources as they come online:

CEO agent
Browser testing bot
Webhooks
Support chats & ticketscoming soon
Error & performance logscoming soon
Usage analyticscoming soon

05

Transparent autonomy

Autonomous doesn't mean opaque. Every change SIMPSEN deploys is individually visible, measured, and reversible.

Each batch is tracked end to end—CEO rationale, dev agent logs, merge results, and browser test feedback. Every change lands on its own branch. If something goes wrong, git revert isolates the blast radius to a single PR.

The dashboard adds a transparency layer—see what SIMPSEN has planned, what each agent built, and how test feedback shapes the next batch.

06

Learn more