November 2026 Summaries

Claude Fable 5.2 is now available on Entire for creating custom review agents within Trails, a pull-request alternative centered on confidence evaluations. Trails allow teams to define requirements for code changes and use configurable agents to assess criteria such as code quality, security issues, and drift on every push. Anthropic reports that Fable 5.2 offers improved coding and reasoning performance, with its High effort setting performing approximately on par with Opus 5.5 at Max effort while costing about half as much per task. The model’s API pricing is $8 per million input tokens and $40 per million output tokens, representing a 20% reduction from Fable 5.1, and it can be selected through Entire’s web interface or configured from a coding agent.

Entire has added RWX as a native CI integration for Trails, its pull-request replacement, expanding its existing options alongside Buildkite, Depot, and Dagger. RWX executes workflows as graphs of independent, cacheable tasks, enabling parallel execution, failure-specific retries, avoidance of unchanged work, task-specific compute sizing, per-second billing, and more detailed results than raw logs. To enable the integration, users install the RWX plugin through an Entire project’s Plugins page, authorize repository access in RWX, confirm the Entire VCS connection and status checks, and add an RWX run definition with an Entire trigger for pushes or Trail updates.

Hermes, an open-source AI agent from Nous Research designed to retain memory across workflows and platforms, can be integrated with Entire to link agent activity to code changes and preserve development context in repository checkpoints. The preview integration requires installing Hermes, the Entire CLI, Go, and a separately built Hermes adapter, configuring an isolated Hermes profile, enabling external-agent support in a fresh Git repository, and restarting the Hermes gateway if necessary. After Hermes creates a file using an absolute path, Entire can verify the uncommitted agent activity through pending checkpoints before the change is committed; the resulting Git commit includes an Entire checkpoint trailer that connects it to captured context. Users can inspect checkpoint details or generate an AI summary after authenticating a configured summary provider. The integration captures sanitized prompts and final responses, tool names and broad statuses, and modified file paths, while excluding Hermes memory, prior conversation history, tool arguments, and raw tool results.

Entire has introduced native Dagger CI integration, positioning continuous integration as an important safeguard for software increasingly written by agents. Dagger, described as a third-generation CI system, can run the same pipelines before and after code is pushed while using shared caching and scalable compute. Its pipelines operate within Entire’s Trails, a pull-request replacement, and report results to configurable Gates that determine the requirements for merging. Users can enable the integration by installing the Dagger plugin for a project, granting repository access, and activating Head and Trails checks in Dagger; pushes to Trails and merges into the default branch then trigger builds. The integration expands Entire’s CI options alongside existing Buildkite and Depot support.

Trails are presented as Entire’s alternative to pull requests for reviewing agent-written code, combining diffs with the coding agent’s captured sessions, checkpoints, timeline, automated review findings, and configurable merge gates. Review agents run independently in cloud sandboxes on matching pushes, assign scores and rationales, and display current results alongside recent runs so teams can assess whether repeated review-and-fix cycles have stabilized. Users can create Trails from branches or planning phases, enable preset agents for code quality, security, and undesirable patterns, or build custom agents such as one that evaluates database migration risks including destructive changes, locking, rollout compatibility, and rollback safety. Trails can span multiple repositories, while GitHub-based repositories may still require linked pull requests and compliance with GitHub branch rules for CI and merging. The workflow emphasizes resolving or assessing findings, pushing updates for rescoring, and merging once review signals are stable and configured requirements are met.