The control layerfor AI-assistedprogramming.

A single smart dashboard that sits above Cursor, Claude Code, and Codex — the central brain for AI-assisted programming. Coordinate agents, route work to the right model, and keep control of cost and change.

Sits above the tools you already use
Claude Code
Cursor
Codex
Antigravity
Scroll
How It Works

One cockpit above every agent

Connect your terminal agents, IDEs and chats into a single workspace. The cost and task router reads each prompt and sends easy work to cheaper models, keeping premium ones for hard problems. Prompt Time Machine snapshots your code before and after every prompt, so each instruction is tied to the exact lines it changed.

OpenPrompt asking which IDEs to capture prompts from
The Problem We Solve

More tools, more models, more chaos

Too much switching costs time. Too many tools and models create chaos. Too many AI changes cost control. What works for one builder does not survive a team or an org. OpenPrompt sits above them to manage that complexity.

01 · Time waste

Context Fragmentation

Too much switching

Developers copy-paste, switch windows, and re-explain the project when they move from Cursor to Claude Code. OpenPrompt keeps one shared context across every agent.

02 · Money waste

Token Sprawl

Too many models and tools

Premium models get used to rename variables and format code. The router reads the prompt and sends easy work to cheaper AIs.

03 · Debugging pain

The Black Box

Too many AI changes

An agent edits 15 files and the build breaks. Prompt Time Machine ties each instruction to the exact lines it changed, so you can see what introduced the bug.

04 · Scale pain

Knowledge Silos

From one builder to the org

A solo builder can keep prompts and context in their head. A team of 2–100 cannot share them, and enterprise has no governance, security, or visibility. OpenPrompt is the layer that scales with headcount.

Integrations

Works With the Tools You Already Use

OpenPrompt plugs into your existing AI workflow — no need to change how or where you code.

01 · Terminal agent

Claude Code — in the same cockpit

Claude Code stays your terminal agent. OpenPrompt is the hub above it — shared context, prompt history, and a before/after snapshot of what it changed.

02 · AI coding environment

Cursor — we sit above the editor, not inside it

Cursor is the AI coding environment. OpenPrompt does not replace it. We coordinate the agents around it, route cheap vs premium models, and keep a history of what AI did.

03 · Coding agent

Codex — one workspace, every agent

Codex is in the same dashboard as Cursor and Claude Code. Context stays with the project, so you do not re-explain the stack every time you switch agents.

Pricing

Individuals, teams, then enterprise

The control layer for AI-assisted programming. One cockpit above Cursor, Claude Code, and Codex — priced for a single seat, a team, or the whole org.

Free

For trying OpenPrompt on a real project before paying for anything.

$0/ mo
no card required
  • 10 Smart Enhance and 10 commit messages a day
  • 20 Quick Enhance calls a day
  • Up to 5 projects, 3 named agents each
  • Full desktop app on macOS and Windows
  • Agent Mode is not included on Free
Start free
Individual

For builders and founders who need prompts, context, and workflow in one place.

$15/ mo
one seat · billed monthly
  • Unified multi-agent hub for Cursor, Claude Code, Codex and Antigravity
  • Prompt Time Machine — before/after on every prompt
  • Cost & task router — cheap models for easy work
  • OpenPrompt Agent on your project context
  • Unlimited named agents per project
  • Local capture, git diffs, and project index
Get Individual
★ Teams
Team

For teams of 2–100 that share workflows, project knowledge, and cost control.

$40/ user / mo
per seat · billed monthly
  • Everything in Individual
  • Shared AI workflows across the team
  • Multi-agent management in one cockpit
  • Shared project context and prompt history
  • Cost control so premium models stay on hard work
Start a team
Enterprise

For orgs that need governance, security, and visibility across AI development.

Custom
custom quote
  • Everything in Team
  • Centralized AI management across the org
  • Governance, administration, and security
  • Cost optimization and usage visibility
  • Built for enterprise engineering at scale
Talk to us

Every limit, every feature, side by side — including what Team and Enterprise add.

Compare all plans
Download

Get the desktop app

OpenPrompt runs on your machine, next to the editors it watches. Install it, link a project folder, and your prompts start filing themselves.

macOS

Apple Silicon or Intel · macOS 13 or later

v0.1.7 · 10.2 MB
Windows

Windows 10 or later · x64

v0.1.7 · 8.2 MB
Linux

AppImage or .deb · x64

Coming soonWatch the releases to hear when it lands

System requirements, install steps, and every published build — in one place.

All download options
FAQ

Questions, answered

What it captures, where it goes, and what it refuses to touch.

Not for capture or storage. Prompts live in SQLite at ~/.openprompt/prompts.db and projects in a local JSON store. Model calls only happen when you ask for Enhance or a commit message, and those try your local agent CLI first.

Claude Desktop on macOS through the Accessibility API, and Cursor, Codex and Antigravity from the chat logs they keep in each project folder. Add that folder to an OpenPrompt project and prompts are filed automatically.

By the folder your IDE has open. Each project owns one or more folders, and the longest matching folder wins, so a repo nested inside another still files correctly. Prompts from folders no project owns are discarded.

A markdown file at .openprompt/OpenPrompt.context.mdc in every folder a project owns — your instructions, recent prompts with the files they changed, and the key files to read. It is refreshed automatically and lives in a hidden directory with its own .gitignore, so it never shows up as a file change or in your commits.

No. Credential-shaped files are dropped unconditionally — env files, .pem and .key files, anything named credentials or containing "secret". Dependency directories, lockfiles, minified bundles, binaries and files over 256 KB are skipped too.

Per project, from its ⋯ menu in the sidebar. Other projects keep monitoring. Resuming re-seeds the change baseline, so the gap is not reported as one giant diff.

Claude's Electron UI can't host an in-app button, so capture and Enhance use the Accessibility API and a listen-only Enter tap. Shift+Enter, chat switches, clears and slash-command picks are ignored. It's required only for Claude Desktop — Cursor and the other IDEs are captured from their on-disk transcripts.

Free covers 10 Smart Enhance calls, 10 commit messages and 20 Quick Enhance calls a day across up to five projects, with three named agents per project and no card required. Individual is $15/month for a single seat. Team is $40 per user per month for shared workflows, multi-agent management, and cost control. Enterprise is a custom quote for centralized management, governance, security, and visibility — talk to us for your seat count.

No. Those stay your coding tools. OpenPrompt sits above them as the control layer — one hub for prompts and context, a router that sends easy work to cheaper models, and a Time Machine that ties each prompt to the lines it changed.