Skip to main content
// ple

// three labs tell you to iterate — none give you the tool

Engineer your prompts.
Don't just save them.

One versioned prompt library for Claude Code, Cursor and Codex. Import the commands you already keep, pull the current version over MCP or the CLI, roll back a bad edit, and test variants against your own agent with Loop.

Get started free
$npm i -g promptlike

Try without an account: version, diff and roll back a prompt in your browser.

tested in Claude Code  ·  setup guides for Cursor and Codex

CLI  ·  Web  ·  MCP  ·  diff + rollback  ·  Chrome extension soon

Example flow: a prompt saved in the terminal with ple save and uploaded with ple sync appears in the web app's Version History as v1.0.0. After two more saves, the diff between v1.0.1 and v1.0.2 shows what changed, and rolling back restores v1.0.1 as a new version, v1.0.3, without deleting anything.

// how it works

One prompt, seven tasks, start to finish.

Follow a single code review prompt from the first ple save to your agents and your team. Every line on screen mirrors the real CLI and web app.

  1. 01ple save

    Save it from your terminal. No account.

    ple save writes a plain Markdown file to ~/.promptlike/prompts and spots its {{variables}}. Show, get, copy and search all work signed out.

  2. 02ple login · ple sync

    Connect an account when you want history.

    Sync is explicit and one-way: it copies, it never moves your files. Each prompt lands as v1.0.0 “Imported”, and every save after that is a new version.

  3. 03{{variables}}

    Write it once, fill it per use.

    In the web app, Fill variables shows a live preview before you copy. In the terminal, ple get --var prints raw text you can pipe straight into an agent.

  4. 04diff · rollback

    See what changed. Undo the bad save.

    Compare any two versions field by field. Rollback restores an earlier one as a new version, so nothing is deleted and the rollback itself can be undone.

  5. 05ple loop --execPro

    Let your own agent pick the winner.

    Loop writes three variants (concise, structured, chain-of-thought) and runs each through the command you already use. You pick; the winner is saved as the next version.

  6. 06promptlike-mcp

    Your agent pulls the current version.

    Your prompts show up in the / menu. Read tools (list_prompts, search_prompts, get_prompt, fill_variables) plus save_prompt / update_prompt when you ask — every write is a new version you can roll back.

  7. 07collaboratorsPro · Team

    Share the library with your team.

    Invite by email, pick a role (viewer, editor or admin) and send the link. The repo appears under Shared with you, with one history for everyone.

// loop engineering

Stop guessing which prompt is better. Loop it.

“Improve” buttons are one-shot and subjective. Loop is an iteration workbench: it generates pattern-based variants, tests them against your own agent, and saves the winner as the next version, so every iteration is recorded, diffable, and reversible.

in the editor
  1. 1.Open a prompt and hit Loop next to Improve.
  2. 2.Optionally tell it what to optimize for, like “fewer tokens” or “stricter output format”.
  3. 3.Compare 3 variants side by side (with token deltas), apply one.
  4. 4.Hit Save and the changelog is pre-filled with the strategy and rationale.
in your terminal
# single round, pick in place
$ ple improve code-review
# multi-round, judged by YOUR agent
$ ple loop code-review --exec 'claude -p {}' --rounds 3

Works with any agent CLI. The variant is passed as PLE_PROMPT, on stdin, and as a {} placeholder.

Read the Loop docsStart looping your prompts →Pro feature

// one source of truth

The same prompt, three surfaces.

Write once. Reach it from your terminal, your agents, or the web, always the same version.

CLITerminal
$ ple show code-review
# Code Review v2.1
You are a senior engineer.
Focus: {{area}}
─ copied to clipboard

ple show, copy, get --var. Raw, agent-readable output

MCPYour agents
claude code
tool suggest_prompts
→ code-review
tool get_prompt
→ code-review v2.1
tool fill_variables
→ area: auth, ready to use

Claude Code reads it as tools, slash commands and @-resources. Always the current version; saves are add-only

WebWeb app
Code Review Promptv2.1
You are a senior engineer.
Review this code for…
v2.0v1.3v1.0

Edit, version, organize, share. Your prompt repository in the browser

// the library at work

Collect, version, connect, measure.

  • 01ple import --scan

    Collect what you already wrote.

    Scans Claude Code, Cursor, Codex and Copilot prompt folders plus AGENTS.md. Identical copies count once.

  • 02ple log

    Every save is a version.

    Semver per save, the source it came from, and a rollback that adds a version instead of erasing history.

  • 03promptlike-mcp

    Plugged into your agents.

    Claude Code pulls the current version over MCP; other MCP clients can use the same server. No copy-paste, no stale duplicates.

  • 04insights

    See what actually gets used.

    Daily uses and where they came from: the CLI, MCP tools, slash commands or the web. Same numbers in ple stats.

// capabilities

Everything you need to engineer better prompts.

cli
CLIple

ple show <name> gives raw, agent-readable output with {{var}} template filling

ver
Versioning

Git-style history, rollback any version, visual diff between saves

loop
Loop Engineeringpro

Generate variants, test them with your own agent, save the winner as a version

pat
Pattern Library

10 copy-ready patterns: Chain-of-Thought, Few-shot, Output contract, with {{var}} templates

cost
Token Cost

Live token count and per-call cost as you write, priced across Claude, GPT, and Gemini

org
Projects + categories

Organize 200+ prompts without losing them: projects, tags, ⌘K search, Recently Used

priv
Private by default

Your prompts stay yours. Public sharing is optional, never forced. JSON export always

mcp
MCP servernpx promptlike-mcp

Your library as agent tools: Claude Code reads, searches, and fills your prompts directly

shipping soon
ext
Chrome extensionsoon

Your library beside ChatGPT, Claude, Gemini: fill {{vars}} and copy without leaving the page

// the old way

Your best prompts shouldn't live in scattered text files.

scattered
notes.txt
Important prompt for SEO... do not delete!!
slack
hey can you send me that prompt again?
Untitled.txt
Act as a senior developer... [v2 final final]
gist
which version worked??

Scattered across tools, lost versions, overwriting work that mattered.

one source of truthmy-library
Product Launch Email
edited 2h ago · v4.2
ready
Customer Support Agent
edited 5h ago · v12.0
draft
Technical Doc Writer
edited 1d ago · v1.1
One library. Versioned. Always structured.

// why ple exists

Skills are starting to sync. Your prompts still scatter.

Claude Code now downloads your claude.ai skills when you sign in. It is one way, and it covers Skills only, not the prompts you reuse across ChatGPT, Claude, Gemini and the terminal.

official documentation

“Custom Skills do not sync across surfaces… Skills uploaded to claude.ai must be separately uploaded to the API.”

Anthropic Agent Skills documentation

The exception: since Claude Code v2.1.273, a terminal session signed in with your claude.ai account downloads that account's skills to your machine. The sync is one-way. Edits you make locally are never uploaded back.

Claude Code skills documentation

Anthropic's ecosystem is excellent. PLE doesn't sync Skills. It keeps your prompts in one versioned place:

+Write a prompt once in the web app
+Pull it from your terminal with ple show
+Use the same prompt beside ChatGPT, Claude, or Gemini (Chrome extension soon)
+Version everything, roll back any time

// agent-native

Your agents read it. You version it.

Context engineering starts with prompts you can trust — and MCP itself has no history, no diff, no rollback. PLE is the record layer underneath: agents pull the current version over MCP, you keep the full engineering history, from CLAUDE.md checklists to review prompts.

promptlike-mcplive on npm
list_promptsbrowse the library, tags included
search_promptsfind by title or description
get_promptfull content, version-pinnable
fill_variables{{vars}} in, ready-to-use text out
suggest_promptsup to 3 of your prompts that fit a task
save_promptadd a new prompt when you askwrite
update_promptadd a new version; the old one is keptwrite
$ claude mcp add promptlike -- npx -y promptlike-mcp

Add-only by design: an agent can browse and fill your prompts, and save a new one or a new version when you ask. There is no delete tool, an update never overwrites the old version, and it can never burn your AI tokens. Same login as the CLI, zero extra setup.

example · cost, before you ship412 tok≈ $0.0012/call · claude−32% after Loop

Live token count and per-call cost in the editor, priced across Claude, GPT, and Gemini, plus a token delta on every Loop variant.

// build log

No testimonials yet. Here are the receipts.

Built by one developer, with no invented reviews and no fake logos. The proof a new tool can honestly show is that its own code ships green and tested.

2524 tests  ·  3 surfaces  ·  MIT-licensed npm packages

$ vitest runpassing
✓ web 2064 tests
✓ cli 326 tests
✓ mcp 134 tests
✓ build 0 errors
→ published: npm i -g promptlike
Test Files  green  ·  counted 2026-10-13

// pricing

Start free. Pay only for the engineering.

Saving and organizing is free. Paid plans unlock the parts a text file can't: unlimited history, AI iteration, and team seats.

Free
$0

Start organizing your prompts without friction.

·Up to 25 prompts
·1 project
·Up to 3 versions per prompt
·Web, CLI, and MCP server
·Chrome extension (soon)
·{{variables}} templates + pattern library
·Live token counter in the editor
·3 AI Improves a month to try it
·JSON, AGENTS.md & Skill export, always
Get started
Prorecommended
$19/mo · $180/yr

Unlimited prompts, unlimited history, no friction.

·Everything in Free
+Unlimited prompts & projects
+Unlimited version history
+AI Improve: one-shot prompt variants
+Loop: test variants with your own agent
+1M AI tokens / month + $3 MicroBoost top-ups
+2 collaborator seats
See Pro
Team
$39/mo · $390/yr

One shared prompt library for your whole team.

$7.80 per member seat with all 5 seats filled. The owner doesn't use a seat.

·Everything in Pro
+Up to 5 member seats
+Members inherit unlimited versions
+3M pooled AI tokens / month
+Role-based access (admin, editor, viewer)
+Per-member usage audit
See Team

// notes

What the labs actually say about prompting.

Notes from primary sources — Anthropic, OpenAI and Google documentation — on versioning, examples and iteration. Cited, short, and no more than a few times a month. No account needed.

We keep your address only to send the notes. You are on the list only after you confirm it by email. To stop, write to zeynepaslierhan@gmail.com and we remove it. See the Privacy Policy.