Alma by Olivares AI

Alma by Olivares AI is an advanced AI assistant designed to solve one of the most persistent frustrations with current AI tools: the lack of long-term memory. Unlike standard chatbots that treat every conversation as a fresh start, Alma automatically extracts and retains facts, preferences, and decisions from every interaction, building a continuously evolving understanding of who you are and what you need. This persistent memory means your AI never forgets your preferred writing style, your ongoing projects, or the specific details you shared weeks ago. Beyond memory, Alma introduces a unified workspace that combines background agents, creative studios, and research tools under a single monthly budget. You can set long-running research agents that work for days or weeks, generate images, videos, and music without juggling separate subscriptions, and conduct deep web research that consolidates findings directly into your memory. Alma is built for everyone from casual users who want a smarter conversational partner to creators producing thumbnails and soundtracks, researchers tracking trends, and power users who want to bring their own API keys or connect via MCP. It supports 16 languages, offers voice conversations and voice note transcription, and provides a long-form editor with export to DOCX, XLSX, PPTX, and PDF. With a starting price of just $14 per month and no hidden add-ons, Alma aims to be the single AI tool that truly remembers and works for you.

Persistent Memory

Alma features a complete memory system that goes far beyond a simple vector store. It automatically extracts facts, preferences, and decisions from every conversation, building a continuously evolving understanding of the user. This system understands context, personality, and learned behavior, ensuring that the AI never forgets your preferred writing style, ongoing projects, or specific details shared weeks ago. Memory is persistent across all interactions, making every conversation more relevant and personalized over time.

Background Agents

Alma allows users to describe a goal in plain language, and the AI plans and executes it in the background for hours or even weeks. Agents can be tasked with long-running research projects, such as investigating a competitor for 30 days and sending daily briefs and a final report. They can also watch the web for brand mentions and notify the user when relevant content appears. Additionally, agents can be scheduled to generate status reports from recent memories on a recurring basis, like every Monday at 8am.

Creative Studios

Alma bundles three creative studios into one workspace: Image Studio, Video Studio, and Music Studio. These studios share a single monthly budget with no add-ons or separate subscriptions. The Image Studio uses Flux Pro and Leonardo across various styles, sizes, and aspect ratios, with features like inpainting, upscaling, and bulk variations. The Video Studio leverages Runway Gen-4 and Gen-4.5 for storyboarding and stitching up to 60 seconds of 1080p video. The Music Studio uses ElevenLabs for studio-grade tracks up to 10 minutes, including focus loops, intros, and demos with or without vocals.

Soul Engine v2

The Soul Engine v2 defines a complete cognitive framework for the AI, not just a personality prompt. It allows users to set an identity, worldview, style guide, anti-patterns, and communication modes. This feature ensures that the AI interacts in a manner consistent with the user's preferences and expectations. Alongside Context Assembly, which automatically assembles the right context for each interaction without manual prompt engineering, Soul Engine v2 makes Alma a truly personalized and adaptive assistant.

For Everyday Users

Casual users benefit from a smarter conversational partner that remembers them between conversations. Alma provides natural chat with persistent context, allowing users to pick up discussions exactly where they left off. Voice conversations on the move and voice notes that are transcribed into structured memories make the assistant accessible and convenient. Whether discussing personal projects or daily tasks, Alma adapts to the user's tone, preferences, and the specific details they care about.

For Creators

Creators can use Alma to generate thumbnails, soundtracks, and short videos without juggling multiple subscriptions. The Image Studio handles multiple models for diverse visual content, the Video Studio produces short video generation in seconds, and the Music Studio creates full tracks up to 10 minutes. All of these creative tools draw from a single monthly budget, simplifying cost management. This unified workspace is ideal for content creators who need efficient, high-quality asset production.

For Research and Knowledge Work

Researchers and knowledge workers leverage Alma for deep web search, trends tracking, and long-form content creation. Web search results are summarized directly into the AI's memory, providing a persistent knowledge base. Trends tracking monitors ongoing topics over time, and the long-form editor supports export to PDF, DOCX, and Markdown. Alma keeps the thread across multiple projects without contaminating each other, making it a powerful tool for academic, market, or competitive analysis.

For Power Users

Power users can maximize Alma's capabilities by bringing their own provider keys (BYOK) for compatible AI models, connecting Alma to their editor over MCP (Model Context Protocol), or hitting the API directly. These features are available on the Max plan. Additionally, a VS Code extension with sidebar chat allows developers to integrate Alma directly into their coding workflow. This use case is tailored for those who need deep customization and integration with existing development environments.

How does Alma's persistent memory work?

Alma automatically extracts facts, preferences, and decisions from every conversation you have. This information is stored in a continuously evolving memory system. When you start a new conversation, Alma assembles the relevant context from your memory to provide a personalized and coherent experience. There is no need for manual prompt engineering; the AI understands what you care about and applies that knowledge across all interactions.

Can Alma run tasks in the background for long periods?

Yes, Alma supports background agents that can run for hours or weeks. You simply describe a goal in plain language, and Alma plans the task and executes it on a schedule. For example, you can ask it to investigate a competitor for 30 days, sending daily briefs and a final report, or to watch the web for mentions of your brand and notify you when relevant content appears.

What creative tools are included in Alma's studios?

Alma includes three creative studios: Image Studio, Video Studio, and Music Studio. The Image Studio uses Flux Pro and Leonardo for generating images with various styles and sizes, including inpainting and upscaling. The Video Studio uses Runway Gen-4 and Gen-4.5 to produce up to 60 seconds of 1080p video. The Music Studio uses ElevenLabs to create studio-grade tracks up to 10 minutes, with or without vocals. All studios share a single monthly budget.

What export formats does Alma support for documents?

Alma provides a long-form editor that supports exporting your work to several common formats. You can export documents as DOCX (Microsoft Word), XLSX (Microsoft Excel), PPTX (Microsoft PowerPoint), and PDF. This makes it easy to integrate your work with standard office suites and share it with others. The editor is designed to keep the thread across multiple projects without contaminating each other.

Alma by Olivares AI starts at $14 per month. This single monthly budget covers all features including persistent memory, background agents, and the three creative studios (Image, Video, and Music). There are no hidden add-ons or separate subscriptions for different features. Users can top up their budget only when it is exhausted, and they can cancel their subscription at any time. A Max plan is also available for power users who want to bring their own API keys, connect via MCP, or use the API directly. All user data is exportable, and the service supports 16 languages.