reduce your agent costs →

ai won't replace humans

ohtools is the open-source framework for explorable, AI-operable TypeScript tool apps.

$npx @bosun-sh/ohtools create my-tools

npx skills add https://github.com/bosun-sh/skills --skill ohtools

philosophy — 001

execution → ownership

bosun removes humans from execution loops and elevates them into decision loops. agents execute. engineers define what matters.

engineer time allocation
yesterdaytodayexecuting70%reviewing20%deciding10%defining20%reviewing20%deciding60%
OBS. 01

decision over execution

engineering value shifts from writing code to defining systems. humans make decisions; agents handle execution within those constraints.

OBS. 02

spec as the source of truth

every dev action traces back to a spec. this turns requirements into executable contracts agents can read. no lost context, no hallucinated intent.

OBS. 03

autonomous feedback loops

agents don't need supervision, they need systems. tests, guards, and metrics create a closed loop where execution is continuously evaluated.

0
projects
0→1
spec to ship
agent hours
0%
human intent
how we work — 002

the harness for autonomy

six stages. one closed loop. scroll to walk the pipeline.

the loop
stage 01spec01spec02assign03strategy04execute05secure06feedback
spec → assign → strategy → execute → secure → feedback
01
spec

capture intent

write specs with kpi and okr. requirements become executable contracts that agents can read and act on.

02
assign

match skills

assign the right agent to the right task. project management meets agentic execution. specialized agents for specialized tools.

03
strategy

establish guardrails

define rules, principles, tech strategy. set the boundaries within which agents operate autonomously.

04
execute

let it fly

run deterministic atomic scripts with integrated tests and linters. give agents tools instead of letting them fumble.

05
secure

stress test

continuous pentesting and cybersecurity scans. adversarial agents probe, attack, report.

06
feedback

iterate

closed-loop feedback. measure what broke, refine the harness, add new rules. the system learns from its own mistakes.

products — 003

the fleet.

seven products. three verticals. platform compounds modules and teams.

modules
$ bosun fleet --modules
M01chest
closed

metrics harness to measure task completeness and impact.

M02plot
closed

spec engineering framework. spec-driven development harness.

M03sounder
closed

depth checks. estimate tasks and measure your agents capabilities.

M04logbook
released · oss

ai-facing kanban. task management tailored for agentic collaboration.

teams
$ bosun fleet --teams
T01crew
in development

diverse agent team tailored for autonomous product engineering.

T02raiders
in development

adversarial agent team. continuous pentest and prevention in your harness.

the stack

platform primitives make modules reusable and teams faster.

primitivesplatformpackage the loopmodulesship at agent speedteams
Team — 004

the team.

founder 01
JR
nameJulian Ramirez
roleFounder
domainAI Engineering

architects the autonomous agent infrastructure behind bosun. background in machine learning systems, platform engineering, and production-scale ai pipelines. builds the systems that build the software.

founder 02
SP
nameSofia Plajutin
roleFounder
domainProduct Security

leads security architecture across bosun. expertise in red team operations, adversarial testing, compliance frameworks, and data governance. security engineered in, not bolted on.

our philosophy
small models outperform large ones for specific agentic work if given the right tools.ai won’t replace humans.
Early Access — 005

join the first wave.

bosun is under active development.
get early access. we ship new stuff every week.

$

Spec-driven · Agent-powered · Human-commanded