Documentation

How to do everything in Prompterjack — from analyzing your first repo to shipping AI-context checks across your team.

What is Prompterjack?Analyze your first repo The GitHub App: PR checks, auto-fix & private repos Teams: shared fleet health & delivery analytics Agent Architect: design & export multi-agent systems System Essentiator: click together a Claude Code setup Pipeline delivery: shareable link or push to your S3 The prompt toolkit More guides & plans

What is Prompterjack?

Prompterjack is the ops platform for your AI coding setup. Point it at a repo (public or private) and it reads your actual code, then generates and maintains the context files that AI coding tools — Claude Code, Cursor, Copilot — rely on: CLAUDE.md, AGENTS.md, a tailored agent roster, and .mcp.json. Install the GitHub App and it keeps them healthy on every pull request. Alongside the codebase tools sits a full prompt toolkit and a multi-agent architecture designer.

The short version:

• Analyze a repo → get a CLAUDE.md built from your real conventions, as a reviewable PR
• Keep it healthy automatically — PR checks + auto-fix PRs when it drifts
• Share your fleet’s health with your team + see delivery analytics
• Design multi-agent systems and export runnable code for 6 frameworks

New here? Start with Analyze your first repo below.

Analyze your first repo

The core flow — from a repo to a ready-to-use Claude Code setup. Free includes 1 codebase; Pro manages a fleet of up to 25.

Add a codebase

Go to Codebase Analysis and paste a GitHub URL, upload a ZIP, or paste a snippet. (No account yet? Paste a public repo on the homepage for an instant health check, no signup.)

We analyze it against ground truth

Your frameworks, types, and conventions are profiled deterministically. If you already have a CLAUDE.md, it’s scored against your real code and stale references are flagged.

Read the results

The Dashboard tab shows your detected stack and an AI-context level (none → thin → partial → rich — how ready your repo is for AI tools). The Recommendations tab lists source-linked, severity-tagged findings.

Get your Claude Code setup

Open the Claude Code tab and download your CLAUDE.md, AGENTS.md, agent roster, and .mcp.json — or, on a connected repo, deliver them as one reviewable PR (see the GitHub App below).

Re-scan as your code changes

Re-scan any time; every run is recorded with severity counts so you can see what changed. On Pro, daily monitors do this for you and alert on drift.

The GitHub App: PR checks, auto-fix & private repos

Install once and Prompterjack keeps your AI context healthy on every push — no manual re-scans.

Connect a repo

Analyze a GitHub repo, then open the GitHub connection card on the codebase page and follow the prompt to install the Prompterjack GitHub App on that repo’s owner (your account or org). Claim the installation to link it to your account.

Analyze private repos (Pro)

Once the App is installed and claimed on the owner, private repositories analyze just like public ones — your code, your permissions. Paste the private repo’s URL and analyze.

Health checks on every PR

The App posts an AI-context health check on each pull request, so context drift is caught in review — not after a bad Claude session.

One-click setup PRs & auto-fix

Use “Create setup PR” to deliver your whole agentic setup (CLAUDE.md, AGENTS.md, agent roster, .mcp.json) as a single add-only PR — hand-written files are never overwritten. Opt into auto-fix PRs and Prompterjack opens a PR to refresh your CLAUDE.md whenever it drifts.

Prompterjack requests least-privilege permissions and delivers changes only as reviewable PRs — it never pushes to your branches directly.

Teams: shared fleet health & delivery analytics

Create a team

Go to Teams (also in your account menu) and create one. On the Team plan, every member gets Pro on one flat bill — not per seat.

Invite teammates

Share an invite link — no email required. Anyone with the link joins your team.

Share a codebase

From a codebase’s page, share it with your team so members can see its AI-context health. Your personal codebases stay private unless you share them.

Read the delivery analytics

The team dashboard shows delivery-analytics signals: how much of your shipping is AI-authored, which tools write it (Claude Code, Copilot, Cursor…), and whether the fixes actually merge. (Signals populate once the GitHub App has observed a few pushes.)

Agent Architect: design & export multi-agent systems

Describe your system

Open Agent Architect (Pro, or bring your own API key) and describe what you want to build, or start from an analyzed codebase. Pick an architecture style — pyramid, sequential pipeline, reflection loop, swarm, and more, including a dedicated Microsoft Copilot Studio structure.

Design & review

Prompterjack designs each agent’s role, tools, guardrails, and memory needs, with an interactive diagram and an eval suite.

Export runnable code

Export to 6 targets: Claude Code, LangChain/LangGraph, OpenAI Agents, Strands, LlamaIndex, and Microsoft 365 Copilot Studio — plus PDF documentation. Then hand it to your pipeline (below).

System Essentiator: click together a Claude Code setup

The System Essentiator builds a complete Claude Code setup from scratch — no analysis required. Toggle the pieces you want (hooks, skills, and repeatable task loops), watch a live preview, and download the whole bundle as a ZIP or copy files one at a time.

Use it when you’re starting fresh; use Codebase Analysis when you want a setup grounded in an existing repo.

Pipeline delivery: shareable link or push to your S3

Hand any generated bundle (a Copilot Studio manifest set, an OpenClaw workspace) to your provisioning pipeline — two ways, from the export view.

Shareable link (Pro)

Click “Create shareable link”. Prompterjack zips the bundle and mints a 7-day URL any CI/ALM pipeline can pull with plain HTTP — no cookie, no login:

bash

curl -L "https://prompterjack.com/api/exports/<id>/download?token=<token>" -o setup.zip
unzip -o setup.zip -d ./agent-setup

Push to your own S3 bucket (enterprise)

Open “Push to your S3 bucket”, enter your bucket + credentials, and Prompterjack signs the upload in your browser and PUTs it straight to your bucket — credentials never touch our servers. Your bucket’s CORS policy must allow PUT from this origin with the authorization, content-type, x-amz-content-sha256, and x-amz-date headers. Works with AWS S3 and S3-compatible stores (R2, MinIO) via the endpoint field.

The prompt toolkit

Prompt Crafter

Turn a rough idea into a production prompt through guided questions and iterative refinement, auto-scored on 5 quality dimensions.

Instruction Set

Build a structured, multi-section rulebook for an agent or chatbot, with priority levels per instruction.

Prompt Editor

Revise an existing prompt with variable management, real-time preview, and an on-demand quality score.

Prompt Rater

Score any prompt against the 5-dimension quality system with a Production-Ready verdict.

More guides & plans

Step-by-step guides

• Professional Handbook — min-max prompts for your profession
• SDK Setup — set up Strands and OpenAI agent frameworks
• OpenClaw Guide — step-by-step workspace setup
• API Setup — bring your own Anthropic API key

Free gives you ~10 generations a month and 1 codebase — no card. Pro is $25/month for the full ops platform (fleet of 25, monitors, GitHub App, setup + auto-fix PRs, Agent Architect). Team gives every member Pro on one flat bill. See the pricing page and the public roadmap.

Need a hand?

Stuck on something, or want a feature? Reach out and we’ll help.