From Zapier for Devs to Powering 90% AI Agents
Co-founders of Trigger.dev discuss their journey through three product versions to find product-market fit, how their async infrastructure positioned them perfectly for the AI agent era, and their vision for the future of computing: programmatic checkpoint and restore.
Trigger.dev provides developers with an SDK to add AI agents directly into their products, handling the complexities of execution, long-running workflows, and reliability. While the company now sees over 90% of its usage from agent-based workflows, the path to this point involved several significant pivots and the accidental perfect timing for the rise of generative AI.
The Journey to Product-Market Fit
The company iterated through three distinct versions of its product before finding success:
-
Version 1: “Zapier for Developers”: Launched early in their YC batch, the initial product was an async background jobs framework. It was positioned for back-office automation, such as managing GitHub repositories or marketing tasks, similar to how internal teams might use Zapier. While a Hacker News launch went well, the use cases were primarily internal-facing.
-
Version 2: Embedded Async Tasks: The team quickly realized the most powerful use cases involved embedding
trigger.devdirectly into customer-facing products. This version shifted focus to an SDK for adding async tasks to an existing application, such as processing a document or encoding a video. However, the developer was still responsible for running the code on their own infrastructure. This version “did okay, but it definitely wasn’t product market fit.” A key realization was that a majority of users (around 60%) already thought Trigger.dev was executing the code for them. -
Version 3: The Full Platform: The final, successful pivot was to provide the SDK, the platform, and the infrastructure. Trigger.dev began executing the code on behalf of the developer. Launched in the summer of 2024, this version immediately found product-market fit, with revenue growing over 30% month-over-month. This shift coincided with the explosion of AI, for which their long-running, async execution model was perfectly suited.
Core Technology: The Future is Programmatic Checkpoint and Restore
The foundational vision for Trigger.dev’s architecture is to democratize a powerful, low-level computing concept: programmatic checkpoint and restore.
“We think that’s really where the future of computing is… programmatically being able to do this type of snapshot and restore mechanism because what you really want to do is kind of freeze the compute where it was at… the whole state of the compute, the CPU, the memory, the file system, and then you want to say whenever something else happens like please resume.”
This approach abstracts away the difficult problem of rehydrating state for long-running or paused tasks. For developers, it functions like an operating system scheduler for the cloud, allowing them to pause a workflow, wait for external input (like feedback from a human or another agent), and then seamlessly resume execution without having to manually manage state.
Riding the AI Wave: Real-World Agent Use Cases
The platform’s architecture was accidentally ideal for the agent era. Long-running, interruptible, and stateful workflows are the essence of how AI agents operate.
- Icon.com: A video ad generation platform. Customers upload product assets, and Trigger.dev orchestrates a long-running workflow to process and classify them. An agent then generates hundreds of video ad variations. The workflow can pause for human-in-the-loop feedback before continuing.
- Magic School: An EdTech company that provides AI tools for teachers and students for tasks like lesson planning and homework assistance. The agent execution and workflows are all handled by Trigger.dev.
- Scrappy Bar: A coding agent that can be connected to a GitHub repository to build features. Trigger.dev manages the entire process of pulling code, executing changes, running evaluations with LLMs, and committing the results.
The Impact of Modern LLMs on Building and Hiring
The recent advancements in models like Opus 4.5 and Claude have had a material impact on Trigger.dev’s operations and strategy.
- Hiring: The team has significantly scaled back its hiring plans for engineers. They estimate that developer productivity has increased 5-10x, reducing the need for a larger team. Their hiring process has evolved to prioritize a candidate’s skill in leveraging AI tools, even conducting a paid “trial day” to observe how they work.
- User Base: The gap between “vibe coders” and traditional software engineers has collapsed. Advanced coding agents empower less experienced developers to use sophisticated tools like Trigger.dev effectively, broadening the user base.
- Internal Development: The team uses agents for far more than just writing code. They are used for complex analysis, benchmarking libraries, reviewing code, and ensuring quality. This allows them to “move a lot of that analysis earlier” in the development process, performing tasks that would have previously taken a week in just a few minutes.
Marketing to Agents and the Open Source Advantage
A forward-looking part of their strategy is to treat LLMs as a primary user of the platform. Being open-source provides a massive advantage here, as the entire codebase, tests, and documentation are available for an agent to learn from. The ultimate goal is to create an end-to-end experience where an LLM can decide to use Trigger.dev, create an account, and start executing workflows without any human intervention.