Mainline

Mainline is a Git-native memory layer that gives AI coding agents access to the engineering intent behind a codebase before they make changes. Rather than letting agents infer why code exists from source alone, Mainline records decisions, abandoned approaches, superseded implementations, reviewer constraints, and explicit risks in a local Git-backed store that travels with the repository via custom refs and notes. Agents using Cursor, Claude Code, or Codex read Mainline context via hooks that auto-inject the current intent snapshot at every session start, requiring no manual retrieval step from the developer. Autonomous rate is approximately 20%: context injection and conflict detection run automatically via hooks; intent recording (start, append, seal commands) requires explicit agent or human initiation through the CLI. Pricing is free and open source under Apache-2.0 for the core CLI and protocol. A team tier with hosted collaboration and policy features is planned but not yet released. Setup is minimal: mainline init configures the repo in a single command, installs hooks for supported agents, and is fully operational in under 5 minutes. Integration coverage includes Cursor, Claude Code, Codex, and any agent that can read the Mainline skill file. The CLI is written in Go and runs on macOS, Linux, and Windows. Security: self-hosted and Git-native by design; all data lives in your own Git remote via custom refs and notes, never in a third-party cloud. No SOC 2 or ISO certifications apply to the open source CLI. Hard limits: team collaboration features (shared Hub, policy controls, PR review integration) are planned but not yet released; the current product is solo developer and early-team CLI only. TCO is $0 for the open source core; future team and enterprise pricing has not yet been published.