The promise of an AI agent is that it can do real work. The reality, for most teams, is a chatbot that needs constant hand-holding. Nebula is betting the gap isn't in the model's intelligence, but in its office setup. The startup is selling a workspace where AI agents get the same infrastructure as a human hire: a persistent computer, a set of tools, and a seat in the team channel [nebula.gg, retrieved 2026].
The bet on persistent infrastructure
Nebula's core product is a cloud-based computer, called a Nebula Device, that runs 24/7 for each agent. This isn't a serverless function that spins down. It's a dedicated environment where an agent can keep files open, maintain state between tasks, and be accessed on-demand. The company argues this persistence is what turns a conversational AI into an autonomous teammate. An agent can be tasked with building a Google Slides deck, and it will work through the steps,generating images, handling failures, and retrying,without a human watching the process. The platform layers on three core primitives: Agents that use tools, Jobs that run them on schedules or triggers, and Miniapps that provide interactive surfaces beyond a chat window [nebula.gg, retrieved 2026].
A wedge into team workflows
Nebula is not trying to sell a single, magical AI. Its wedge is the shared workspace itself, a single billable environment where a team can "hire" multiple specialized agents. The go-to-market motion appears designed for bottom-up adoption within technical teams. It offers a free workspace, charges $15 per monthly team seat (with the first two free), and sells compute credits for agent usage [nebula.gg, retrieved 2026]. The early traction signal, an 18% increase in expansion MRR driven by upgrades to the Growth plan, suggests teams that start using it are adding more agents or compute [nebula.gg, retrieved 2026]. The product seems built for the workflows of software engineering, growth, and support teams, where repetitive tasks like code reviews, dashboard updates, and customer ticket triage are prime for automation.
The pricing and procurement puzzle
For a product aiming to become a team's AI workforce, the pricing model is a critical piece of the procurement puzzle. Nebula uses a hybrid structure:
- Seat-based access. Team members pay $15 per month, which covers their access to the workspace and channels.
- Consumable compute. Agents run on Nebula Devices, billed from $0.025 per hour, and use plan credits for Nebula's own models.
- Tooling passthrough. Costs for using external models or API calls are billed separately, not covered by credits [nebula.gg, retrieved 2026].
This creates a predictable cost for human access but a variable, usage-based bill for the AI labor. For a finance team, forecasting the monthly spend for a department of ten agents could be complex. The bet is that the productivity gains will justify the opaque cloud bill, a familiar trade-off in DevOps but a newer concept for general business teams.
Where the concept meets reality
Nebula's ambition is large, but the competitive set is already forming. The company is not competing with ChatGPT; it's competing with workflow automation platforms that are adding AI layers. Realistic alternatives include Zapier's recently launched AI agents, no-code platforms like Make that are integrating AI steps, and specialized agent platforms like Lindy.ai that focus on personal assistants. Nebula's differentiation rests on the dedicated computer and the integrated team channel,positioning the agent as a persistent resource, not a transient tool. The ideal customer profile here is a technical team lead, likely in a startup or mid-market tech company, who is already using Slack, GitHub, and a suite of SaaS tools and is frustrated by the gap between AI demos and automated work. They have the budget to experiment with a new line item for "AI ops" and the authority to bring a new platform into their team's daily rhythm.
The next twelve months will test whether teams are ready to manage AI as infrastructure. Success means moving beyond one-off automations to having agents reliably own slices of operational work. The risk is that the platform becomes another cost center for experiments that never fully graduate to production. For now, Nebula has a clear answer to why your AI agent feels like an intern: it never got its own desk.
Sources
- [nebula.gg, retrieved 2026] Nebula · The AI agent platform for teams | https://www.nebula.gg/
- [nebula.gg, retrieved 2026] Nebula Devices documentation | https://docs.nebula.gg/
- [nebula.gg, retrieved 2026] Nebula blog post on Google Slides automation | https://www.nebula.gg/
- [Crunchbase, retrieved 2026] Nebula - Crunchbase Company Profile & Funding | https://www.crunchbase.com/organization/nebula-333c