/dev/agents Review, Pricing & Alternatives (2026)

/dev/agents began in 2024 under founders David Singleton, Hugo Barra, Ficus Kirkpatrick, and Nicholas Jitkoff. Public coverage described Singleton as a former Stripe chief technology officer and Google Android engineering leader, Barra as a former Google Android product executive and Meta Oculus leader, Kirkpatrick as a former Android engineer and Meta augmented-reality executive, and Jitkoff as a former ChromeOS designer. The company said it was developing a cloud-based operating system for trusted AI agents. Its stated goals included new interface patterns, a revised privacy model, and a developer platform for agent applications. SiliconANGLE reported a $56 million seed round in November 2024. In March 2026, the project emerged from stealth under the Dreamer name, making /dev/agents primarily a historical identity.

Dreamer was presented as a consumer service for discovering, building, and using AI agents and agentic applications. Its main interface was a personal Sidekick that users could name, customize, and address in natural language. The Sidekick could help users find community-built agents, modify existing experiences, and create new applications through conversation. The reported beta included a gallery of agents, an Agent Studio, and tools built by the company or outside developers. Demonstrated examples included calendar research, news monitoring, daily briefings, conference scheduling, and information management. Latent Space reported that the product used a waitlist during its beta period. The company does not publish a current product manual or independently verified service documentation.

The technical proposition extended beyond prompt templates. Latent Space described an SDK, logging, observability, prompt management, databases, storage, serverless functions, and hosted virtual machines capable of running arbitrary code. The platform also included tools and a gallery where community members could publish agents or integrations. Singleton said tool builders could receive payments based on usage, while premium tools could charge users per use after a trial. A reported demonstration showed the Sidekick producing a conference application from structured event information. Another account said builds could take about 10 to 15 minutes because the system iterated, tested, and corrected output. That timing was an interview statement rather than a performance guarantee. The company does not publish a post-transition release schedule.

No current consumer subscription, developer plan, API rate card, or hosted-runtime price is published by the company. During the Dreamer beta, prospective users were directed to a waitlist. The company described free trials for premium tools and usage-based charges for certain tools, including a web research integration, but did not provide a general tariff. Tool builders were told that usage could generate payments, yet no payout formula or settlement terms were published. In September 2026, the public site lacked product documentation, a signup flow, deployment instructions, and pricing. The reported $56 million financing round describes company funding rather than customer pricing. Historical promotional incentives, including a reported $10,000 builder prize, should not be treated as current purchasing options.

The project was aimed at a different audience from several developer-focused agent products. Dreamer emphasized a consumer Sidekick, a community gallery, and conversational creation, while OpenAI AgentKit, Google Agent Development Kit, and Microsoft Copilot Studio present more explicit developer or business entry points in their respective product materials. Replit Agent is a closer comparison for natural-language software creation, while Salesforce Agentforce addresses business agent deployment. Dreamer's strengths are broad platform scope, hosted code execution, community publishing, and tool-builder incentives. It also supports concerns about beta access, incomplete public documentation, unclear pricing, and uncertain continuity after the reported Meta move. Treat the project as a subject for research, not a confirmed procurement option.