Introducing Montage: Remote Config for AI Prompts

Every team building with AI has the same problem: prompts live inside your code.

They're scattered across files, buried in string literals, and changing them means a deploy. A deploy means a PR, a review, a CI pipeline, and a 15-minute wait to find out whether your new system prompt actually made the support bot sound less robotic.

We built Montage to fix this.

What is Montage?

Montage is a prompt management platform for teams. Think of it as remote config, similar to LaunchDarkly or Firebase Remote Config, but purpose-built for AI prompts.

You import your existing prompts, edit them in a collaborative dashboard, and publish changes that your application picks up instantly via our SDK. No redeploy needed.

The Core Loop

The workflow is simple:

1. Import. Paste code, upload files, or connect your GitHub repo. Montage detects your prompts automatically.
2. Edit & Collaborate. Multi-message editor with variables, version history, comments, and approval workflows.
3. Publish. Hit publish. Your app fetches the new prompt via API. Rollback is one click away.

Every publish creates a versioned snapshot with a full diff view, so you always know what changed and can revert instantly.

Built for Teams

Prompts are a product decision as much as a technical one. Your PM wants to tweak the onboarding flow's tone. Your support lead wants to add escalation instructions. Your marketing team wants to update the chatbot's personality.

None of them should need to open a pull request.

Montage gives PMs, marketers, and support teams a safe way to update AI behavior. Approval workflows ensure nothing goes live without a review. Comments and @mentions keep the conversation in context. And version history means there's always a safety net.

GitHub Integration

If your prompts already live in code, Montage meets you where you are:

• Scan & Import. Connect your repo and Montage finds every hardcoded prompt with confidence scores.
• Auto-Replace PRs. One click creates a PR replacing hardcoded prompts with SDK calls.
• Sync Detection. Know instantly when source files change.
• Push Back. Edit in Montage, push changes back to GitHub.

The SDK

Fetching a prompt is 3 lines of code:

import { Montage } from "@montage-sh/sdk";

const montage = new Montage({ apiKey: "mt_live_..." });
const prompt = await montage.get("customer-support-agent");
const compiled = prompt.compile({ userName: "Alex", issue: "billing" });

Variables like {{userName}} are compiled at runtime with type-safe keys. The SDK supports TypeScript and Python, with cURL for everything else.

The CLI

Prefer your terminal? The Montage CLI lets you pull prompts to your local environment, edit in your IDE, and push changes back:

montage pull --project acme-ai
# Edit prompts in your IDE...
montage push
montage publish customer-support-agent --message "Warmer tone"

It works with Claude Code, Codex, Cursor, and any AI coding tool that can run terminal commands. Your AI engineer can manage your AI prompts.

Pricing

Montage has a generous free tier: 3 projects, 10 prompts each, and 1,000 API calls per month. Pro is $10/month for teams that need more. Organizations start at $25/month with advanced access controls and analytics.

Get Started

You can import your first prompt in under a minute. No credit card required.

If your AI prompts live in your code today, they don't have to tomorrow.