HN
Today

Launch HN: Twill.ai (YC S25) – Delegate to cloud agents, get back PRs

Twill.ai launches a new platform that lets developers delegate coding tasks to cloud-based AI agents, promising to automate everything from bug fixes to new features. By running models like Claude Code and Codex in isolated cloud sandboxes, Twill tackles common local agent pain points like parallelization and persistence. Its unique approach to leveraging existing model CLIs and integrating with dev workflows makes it a compelling solution for scaling engineering output.

15
Score
2
Comments
#6
Highest Rank
2h
on Front Page
First Seen
Apr 10, 6:00 PM
Last Seen
Apr 10, 7:00 PM
Rank Over Time
612

The Lowdown

Twill.ai, a Y Combinator-backed startup, introduces a novel approach to developer productivity by enabling the delegation of coding tasks to cloud-based AI agents. Founded by Willy and Dan, Twill aims to resolve the significant hurdles developers face when attempting to integrate AI coding assistants locally, specifically addressing issues of parallel execution, task persistence, and system trust.

  • Twill hosts leading coding CLIs (e.g., Claude Code, Codex) within isolated cloud sandboxes, allowing developers to assign work via various interfaces including Slack, GitHub, Linear, a web app, or CLI.
  • The platform automates the creation of pull requests (PRs), code reviews, diagnoses, and asks clarifying questions, only looping in the user when direct input is required.
  • It solves the limitations of local AI agents such as the inability to run multiple tasks concurrently, agents stopping when a laptop closes, and security concerns over local filesystem access.
  • The system has evolved to include multiplayer capabilities for team collaboration, memory for persistent instructions, and automation features like crons for recurring tasks.
  • Twill's architecture emphasizes using lab-native CLIs, which allows it to automatically benefit from model improvements and offers flexibility by avoiding vendor lock-in.
  • Each task executes in a dedicated sandbox, complete with repo cloning, dependency installation, and process isolation; sandboxes are snapshotted for efficient warm starts.
  • Twill is also open-sourcing agentbox-sdk, an SDK designed for running and interacting with agent CLIs across different sandbox providers.
  • The platform offers various pre-built automation templates for common engineering workflows, including error triage, dependency updates, sprint automation, and documentation drift detection.
  • Pricing includes a free tier providing 10 credits per month, with paid plans starting at $50/month for 50 credits, and a free pro tier available for open-source projects.

In essence, Twill.ai positions itself as an invaluable tool for engineering teams looking to enhance their efficiency and scale by offloading repetitive coding and maintenance tasks to intelligent, cloud-managed agents, thereby allowing human developers to focus on higher-level architectural and product decisions.