Regenerative AIa startup founded and led by former DeepMind founding investor Humayun Sheikh, Release announced today One of three interconnected products designed to provide the trust, coordination, and interoperability needed for a large-scale AI agent ecosystem.
Included in the launch ASI: Onea Personal-A Orchestration Platform; Recover businessan authentication and discovery portal for brand agents; And Agentan open directory hosting over 2 million agents.
Together, the system positions are brought together as an infrastructure provider called the “Agent Web”.
The company says the tools highlight a central limitation in current consumer AI: models can provide recommendations but cannot reliably execute multi-stakeholder actions that require business coordination. The retrieval approach centers on enabling agents from different organizations to securely interact, using authenticated identities and shared context to complete end-to-end workflows.
In a press release provided to VentureBeat, Humayun Shaikh, founder and CEO of FiqqA, and an early investor in DeepMind, said, “We’re building the same foundation for agents that Google created for websites.” “Instead of just searching for information, your personal AI integrates with certified brand agents to do the work.”
Background: Founder of Recovery and DeepMind Connection
Fiqq AI was founded in 2017 by Humayun Shaikh, an entrepreneur whose initial investment in DeepMind helped support the company’s commercial growth prior to its acquisition by Google. “I was one of the first five people in DeepMind and its first investor. My check was the first,” Sheikh reflected on that period.
His early experience helped shape the direction of recovery. “Even in 2013, it was clear to me that agentic systems were going to work. That’s where I focused — on the agent web,” notes Shaikh. Build on this thesis by developing an infrastructure for autonomous software agents, focusing on verifiable identity, secure data exchange, and multi-agent coordination.
Over the years, the company has expanded to a team of 70 people across Cambridge and Menlo Park, raised nearly $60 million, and accumulated more than a million users interacting with its model—which informed the design of newly launched products.
The decision to bootstrap the company initially came directly from DeepMind’s exit proceeds, Sheikh added, noting in the interview that while the sale to Google was “a good way to go,” he believes the team could hold out for a higher price.
Early self-funding funding allowed work to begin in 2015—before the Transformer architecture became mainstream—on the assumption that agentic infrastructure would become the foundation for applied AI.
ASI: A Platform for Multi-Agent Orchestration
is at the core of the launch ASI: Onea language model interface specifically designed to coordinate multiple agents rather than solving isolated queries. Fetch describes it as an “intelligence layer” that handles context sharing, task routing, and priority modeling.
The system stores user-level signals such as preferred airlines, dietary restrictions, budget limits, loyalty program identifiers, and calendar availability. When a user requests a complex task – such as planning a trip with flights, hotel and restaurant reservations – ASI: One assigns those preferences and delegates work to the appropriate certified agents. Agents then return actionable outputs, including inventory and booking options, rather than generic recommendations.
In practice, ASI: functions as a workflow generator across organizational boundaries. Unlike traditional LLM applications, which often rely on APIs or RAG techniques to surface information, ASI: is designed to integrate autonomous agents that can complete transactions. The researchers note that personalization improves over time as the model collects structured preference data.
Sheikh emphasized the difference between orchestrated execution and traditional AI output. “It’s not looking at separate options and hoping they’ll work together,” he said. “It’s orchestration.”
He added that the retrieval architecture is deliberately modular: “Our architecture is a mix of agent and expert models. A big model is not enough – you need experts. That’s why we built ASI1, specifically designed for agent systems.”
The interview also revealed new details about ASI: One’s Personalization System: The platform uses multiple user-owned knowledge graphs to store preferences, travel history, social connections and contextual constraints.
These knowledge graphs are on a per-user basis and are not shared with any retrieval-driven data. Sheikh describes it as a “deterministic backbone” that provides personal AI with a stable memory layer beyond the probabilistic output of a single large model.
ASI: One launched in beta today, with a wider release planned for early 2026. Recovery also offers ASI: Mobile, released earlier this year, giving users access to the same agent orchestration capabilities on iOS and Android. The mobile app connects directly to agent and user knowledge graphs, enabling on-the-go task execution and real-time interaction with registered agents.
Recovery business – verified identity and brand control
To enable trusted synchronization between users and companies, Retrieval is introducing an authentication and discovery portal called Retrieval Business.
The platform allows organizations to verify their identity and claim an official brand agent handle—for example, @Hilton or @Nike—regardless of which tools they use to build the underlying agent.
ICNN domain registration and SSL certificate system-compliant product retrieval for websites. Verified status is intended to protect users from interacting with fake or untrusted agents, a problem the company has cited as a major barrier to mass agent adoption.
The system includes low-code tools for small businesses to create agents in a few steps and integrate real-time APIs such as inventory, booking systems, or CRM platforms.
“With Recovery, you can create an agent in a minute. It gets a handle, like a Twitter username, and you can completely personalize it — even give your social media permission to post on your behalf,” Shaikh said. Once a brand claims its namespace, its agent becomes discoverable to users within the agent’s AIS and to other agents.
The company has pre-booked thousands of brand-name spaces in anticipation of demand. Authentication status is maintained across any platform that the agent integrates with, creating a portable identity layer for business agents.
The interview highlights that the recovery business directly inherits the web trust archetype: domain owners verify their identity by inserting a short code snippet into the backend of their existing website, which allows the system to pass a cryptographic challenge and gives the agent a token of authenticity, like a “blue check” for the agent’s identity. Sheikh coined it as “reusing the trust layer the web has already spent decades building.”
Companies can now start claiming agents Business.Fetch.E.
Agents – An open directory of over 2 million agents
The final component is the release Agentan open directory and cloud platform that hosts agents and enables cross-ecosystem discovery. Millions of agents have already registered, spanning travel, retail, entertainment, food service, and enterprise categories, the recall said.
The agentors provide metadata, capability descriptions, and routing logic that ASI: A uses to identify agents appropriate for specific tasks. It also supports secure communication and data exchange between agents. The company notes that the directory is platform-agnostic: agents built with any framework can be embedded and interoperable.
According to Sheikh, the lack of a discovery layer is one reason most AI agents see little or no use. “Ninety percent of AI agents are never used because there is no discovery layer,” he said.
He formulated the agent’s role in more technical terms: “Right now, if you create an agent, there’s no universal way for others to discover it. That’s what Agentores has solved — it’s like DNS for agents.” He also described the system as an essential component of the emerging agent economy: “Recovery is building the Google of agents. Just as websites need search, agents need discovery, trust and interaction.”
The interview further pointed out that the agent is cloud-agnostic by design. Sheikh linked this competing agent ecosystem to specific cloud providers, arguing that a universal registry is only viable if independent of proprietary cloud environments. The open architecture enables LLM to query any agent “within a minute of deployment,” turning publishing an agent into a process similar to registering a domain, he said.
Agentores also integrates payment channels, enabling agents to process purchases using partners such as Visa, Skyfire, and supported stablecoins. Users can set spending limits or require express approval for transactions.
Industry context and implications
The launch of Fetch comes at a time when users of AI platforms are looking for a shift from static chat interfaces to autonomous agents that are able to complete actions. However, most agent systems are limited by siled architectures, limited interoperability, and weak authentication standards.
Retrieves its infrastructure as a response to these limitations by providing a cross-platform coordination layer, identity system, and directory service. The company posits that an agent ecosystem necessarily requires authentication mechanisms to ensure that consumers interact with authentic brand representatives rather than imitations. By establishing namespace control and portable trust indicators, retrieval businesses aim to fill a gap similar to early web domain authentication.
At the same time, ASI: seeks to centralize user preference data in a way that enables more efficient personalization and multi-agent coordination. This approach differs from generalist LLM applications, which often lack a consistent preference architecture or direct access to brand-controlling agents.
The interview also highlighted that micropayments and digital transaction infrastructure are central to the long-term vision of recovery. Sheikh cited Coinbase’s integration with protocols such as 402 and AP2, and described these capabilities as essential for autonomous agents to complete end-to-end tasks that include financial processing.
Takeaway
ASI’s joint release of Retrieval: One, Retrieval Business, and Agent introduces an integrated stack designed to support large-scale deployment and use of AI agents. The company builds this system as the core infrastructure for an agentic ecosystem, where consumer AIs can coordinate with certified brand agents to complete tasks reliably and securely. Its addition of identification, discovery, and orchestration layers reflects a longstanding thesis of retrieval.