Your AI agent has amnesia. Make the repository its memory.

Description and use

Agent Runway stores decisions, checkpoints, and run logs in the repository, giving AI agents long‑term memory and improving code continuity.

Problem

AI agents forget architectural decisions and mistakes between sessions, causing repeated explanations, drift from requirements, and lack of traceability.

Outcome

Agent Runway enables AI agents to retain architectural decisions, lessons, and checkpoints in the repository, reducing repeated explanations, preventing drift, and providing traceability across sessions.

Workflow

• Initialize Agent Runway with a preset (e.g., vibe-lite) using npx @adonai-labs/agent-runway init --preset vibe-lite --target all
• Use skills such as @express for small changes, @checkpoint to log state, and @learning to capture lessons in .agent-runway/memory
• Optionally upgrade to Structured workflow with npx agent-runway upgrade --to structured or init with node-typescript preset
• Follow the core path: /spec-creator → /ticket-creator → /lead → /review for full implementation with quality gates
• For autonomous runs, use /autonomous-lead which enforces run logs, ADR gates, and human approval before destructive actions
• Use /review in compact or isolated mode to generate PR-style reports with findings and merge gates

Evidence

AI agents are good at writing code. They are not good at remembering what they decided last week.

Source

The only context that survives every session is the repository, so I started treating it as the agent's memory.

Source

Agent Runway is artifact-driven. Work moves through four layers: Spec -> Workflow -> Rules -> Memory.

Source

npx @adonai-labs/agent-runway init --preset vibe-lite --target all

Source