Robyn: an army of AI agents to manage your life while you sleep

Robyn orchestrates AI agents capable of managing emails, research, travel, and professional tasks. Its founder promises 20 hours saved per week.

A simple instruction to delegate emails, prepare for a meeting, search for an apartment, or monitor flight prices, without having to supervise every step. This is the workflow offered by Robyn, a platform of personal AI agents designed by Jenny Shao. Accessible via a web browser or messaging, it allows users to set up multiple autonomous assistants, each assigned to a specific task and capable of carrying out its work outside of direct interactions with the user.

The company promises time savings of over 20 hours per week per person. An ambitious figure, which is nevertheless not accompanied by any public methodology to verify its representativeness. A team of agents from a single instruction

Robyn's interface is built on one idea: describe the desired outcome rather than building a sequence of automations yourself. From a request phrased in natural language, the platform can create a dedicated agent, provide it with the necessary context, and organize its work across various connected services.

The examples presented on the official website cover both personal and professional activities. An assistant specializing in real estate scours multiple listing platforms, compares properties according to specified criteria, and puts together a selection. Another analyzes an inbox, distinguishes urgent messages from secondary communications, and drafts responses. Agents can also monitor airfares, track competitors, prepare presentations, or analyze financial documents.

These tasks are not limited to one-off executions. Robyn allows users to schedule daily, weekly, or monthly tasks. Agents can thus conduct their research overnight and deliver their results at a set time. The user is then presented with a report, the gathered information, and any actions requiring their approval.

The system also allows instructions to be modified after an initial run. When a user specifies a new criterion, the agent can resume its work while maintaining the previously established constraints, rather than restarting the conversation entirely. A memory inspired by neuroscience

Memory is one of the platform's key technical arguments. A former physician, Jenny Shao worked for three years alongside Eric Kandel, a neuroscientist awarded the Nobel Prize in Physiology or Medicine in 2000 for his research on the biological mechanisms of memory.

This background influences Robyn's design, whose architecture is presented as being based on a "neuro-mimetic" memory. The system is designed to retain relevant information about the user, their preferences, habits, and decisions made over the course of interactions, and then supply this context to agents assigned to new tasks.

In a real estate demonstration, a search initially focused on price and location gradually evolves to take into account noise, sunlight, and the presence of a terrace. The new requests build upon the previous criteria, allowing recommendations to be revised without rebuilding all the instructions.

This approach builds on an earlier direction for Robyn. In November 2025, TechCrunch covered the project as a conversational assistant equipped with personal memory and designed to better understand its users. The new platform expands this logic to task execution and the coordination of multiple agents. However, the technical details of its memory system and the benchmarks used to measure its performance have not been published. Over 300 services to interact with everyday tools

The autonomy promised by Robyn relies heavily on its ability to interact with the applications used by its customers. The company claims 336 connected services, including Gmail, Google Calendar, Slack, Notion, GitHub, Linear, Stripe, Shopify, Airtable, Vercel, and Supabase.

Agents can thus operate across multiple environments during a single task. An assistant tasked with preparing a meeting can, for example, search for information in documents, review professional communications, and generate a summary. Another can monitor changes to a software project through GitHub and issue-tracking tools.

The platform also supports operation via iMessage. The user can send a request via text message or continue a conversation with their assistant, and then receive results, documents, and notifications directly within the message thread.

However, this operational capability raises the question of action control. On its website, Robyn states that it asks for confirmation before performing operations deemed risky. Demonstrations notably distinguish between drafted responses and messages that are actually sent. Still, the criteria defining a sensitive action and the security mechanisms applicable to each service would benefit from more precise documentation. Invite-only access and usage-based pricing

For this new version, Jenny Shao announced on X the opening of invite-only access, accompanied by 1,000 codes distributed during the presentation. The founder is also offering 100 free credits to people participating in her launch campaign.

The pricing plans displayed by Robyn start at $30 per month for Starter, with 100 credits renewed monthly and two concurrent runs. The Pro subscription costs $100 per month for 400 credits and up to five parallel runs. The Max and Business plans reach $200 and $1,000 per month, respectively, with higher credit volumes and execution capacities.

Credits are used to cover interactions and the work performed by agents. Consumption varies based on the operations carried out, as illustrated in the platform's demonstrations. A complex research task can thus consume more resources than a simple information query.

Access remains limited during this initial phase. Published demonstrations showcase dozens of use cases, but they do not constitute an independent validation of performance, result reliability, or the claimed average time savings. From conversational companion to task-executing assistant

Robyn's evolution reflects a broader shift in AI usage. Conversational assistants were initially designed