Under the hood
How Nairo works
Nairo passively captures your coding context and distills it into durable memories that get smarter over time. Here's how the engine works from signal to retrieval.
The memory pipeline
Raw signals flow through four stages to become retrievable, high-quality memories.
Observations
Raw signals captured from every tool call as you code
Sessions
Groups observations into coherent sessions with AI-generated summaries
Events
Distilled, durable memories — decisions, conventions, preferences
Context Packs
Relevant memories surfaced back to your AI tool on demand
What happens automatically
Behind the scenes, Nairo continuously processes your memories to keep them clean, relevant, and fast to retrieve.
Deduplication
Near-identical memories are detected and merged at write time, keeping your memory clean without manual curation.
Confidence decay
Stale, unreinforced memories gradually fade over time so old context doesn’t drown out what matters now.
Pattern detection
Recurring patterns across projects are automatically promoted to global scope so every project benefits.
Compaction
Related memories are periodically synthesized into concise summaries, reducing noise while preserving signal.
Hybrid search
Full-text keyword matching and semantic vector search are fused together so you find memories by exact terms or concept.
Reinforcement
Memories you use often grow stronger automatically, surfacing frequently needed context faster over time.
Ready to try it?
Set up once, and your AI tools remember everything from that point on.