ml0x.com/pipeline
The core of this is free, and I would rather you knew that on the way in
The pipeline source is MIT and published at github.com/theluckystrike/100xagenticdev. You can clone it right now for nothing. What is on this page is the same system packaged and pinned at a version, with the configuration, the runbook and the written research that explain why it is built this way. Ninety nine dollars, paid once.
What is in the archive
$99 once. 48 files, one zip, no licence key, no account, no phone home.
bin/100x
The CLI, one file of bash, about 1,160 lines. Four stage run, quality gate, parallel task mode, project init and a doctor command that tells you what is missing before you start.
pipeline/
The parallel research runner and its orchestrator, the model client, the task templates, the result aggregator and the web grounding layer. Eight presets ship with it, from market scan to security audit recon, plus a fan out mode that sends one prompt through twenty perspective lenses.
config/ and scripts/
The hook settings, the NASA Power of 10 eslint config, the MCP setup script, and the quality gate, pipeline and parallel agent runners. That is the part that took longest to get right, and the part least worth writing from scratch.
docs/ and docs/reports/
The instructions, the runbook and the project CLAUDE.md, plus thirteen written pieces of research and architecture, in markdown and HTML. The agentic developer report, the pipeline research report, the deployment strategy and the audit pipeline architecture.
Two ways in
For the Claude Code side, the four stage run with the gates and the budget ceiling.
ln -sf "$(pwd)/bin/100x" /usr/local/bin/100x 100x doctor 100x run "Add input validation to all API endpoints"
For the research side, many agents against one topic, on a ceiling you set.
cd pipeline && pip install httpx DEEPSEEK_API_KEY="sk-..." python3 runner.py \ --preset market_scan --topic "your topic" --agents 20 --budget 5.0
You bring the API key and you pay your own inference. The budget ceiling exists because the bill lands on whoever runs it, and it landed on me first.
What I checked before packaging it
Every run output directory was left out, because those are my own results and not yours, and because generated result files carry placeholder credentials that would trip any scanner you run. Every business specific job config was left out for the same reason.
Two documentation claims were rewritten on the way in. One runaway agent loop figure survived because there is a real published postmortem behind it, and the citation now travels with the number. A second figure and a failure rate statistic did not survive, because no source could be found for either, so they are gone rather than softened.
What this is not
- Not a subscription, and not a hosted service. It runs on your machine against your keys.
- Not a guarantee of a revenue, ranking or time saving outcome. The archive contains no benchmark claims because I ran no benchmarks.
- Not a substitute for reading the code. It is bash and Python and you should read it before you point it at anything you care about.
- Not finished. Nobody except me has completed an end to end run of every stage.
No buyers yet. This went on sale today, so there are no testimonials on this page and there will not be invented ones.
Ninety nine dollars, once
Payment runs through Stripe. You land on a download page immediately, the archive is yours to keep on every machine you own, and the MIT rights to the code stay yours whatever else happens.
Buy the ml0x Pipeline, $99Stripe checkout opens. Card, Apple Pay or Google Pay. Then a download page.
Sixty day refund, no questions asked, and you keep the files. If you would rather clone the free core and skip this, do that instead, the link is in the archive and on the homepage.