Twill

Twill is an autonomous coding agent platform backed by Y Combinator that ships bug fixes, features, and pull requests while developers focus on architecture and product decisions. It follows a structured pipeline — research, plan, implement, test, and PR — ensuring agents can't skip steps, making the process reliable and auditable. Twill integrates with GitHub, Linear, Slack, Sentry, and cloud providers, working where teams already collaborate. It supports multiple AI coding agents including Claude Code, OpenCode, and Codex, and can run them in parallel or repeatedly to maximize success rates.

• Structured Workflow Pipeline: Every task follows a fixed Research → Plan → Approve → Implement → AI Code Review → Merge PR pipeline, preventing agents from skipping critical steps.
• Isolated Sandbox Environments: Agents spin up isolated dev environments to build, test, and iterate on changes, making PRs trustworthy without any manual setup.
• Multi-Agent Support: Choose from Claude Code, OpenCode, and Codex CLI; run multiple agents in parallel or the same agent multiple times to compare outputs and increase success rates.
• Proven Automation Templates: Start from pre-built templates for common workflows like error triage, dependency updates, sprint automation, flaky test detection, and security vulnerability scanning.
• Tool Integrations: Mention @twill from GitHub, Linear, or Slack to assign tasks; SSH into sandboxes to debug with Cursor, VSCode, or any favorite IDE.
• Enterprise Self-Hosting: Deploy Twill in your own VPC with Docker or Kubernetes, SSO/SAML support, full audit trails, and bring-your-own API keys for complete data sovereignty.
• Recurring Workflow Automation: Schedule recurring engineering workflows (daily, weekly, hourly) to automatically triage issues, update dependencies, and keep the PR queue healthy.
• Human-in-the-Loop Control: You approve the implementation plan before any code runs, and you merge the PR when ready — Twill only pings you when it genuinely needs input.