Particle.news

Thomas Dohmke’s Entire Launches Git Forge That Replaces Pull Requests With AI 'Trails'

The product shifts code review to configurable AI evals while storing agent session memory to reduce repeat model costs.

Overview

  • Entire publicly launched its full Git forge on Monday, Oct. 26, 2026, introducing Trails, a dashboard of AI-agent evals that score every push and can permit merges without line-by-line human review.
  • Trails runs multiple agent checks per update and shows customizable scores and confidence levels so teams can choose to require human sign-off, passing tests, or automatic merge rules.
  • The company also unveiled Continuous Memory, which syncs past agent sessions and a codebase graph into the repo so agents can retrieve prior context and Entire says one component cut token use by up to 71% in its testing.
  • Entire offers native private hosting in the U.S., Europe, India and Australia, claims higher push throughput versus incumbents based on internal benchmarks, and is partnering with Buildkite and Depot to migrate CI from GitHub Actions.
  • The launch addresses rising strain from AI coding agents on existing forges but its performance, token-savings and safety claims come from company benchmarks and have not been independently verified, leaving questions about auditability and reliance on automated evals.