Snipara

About Snipara

Snipara is a shared project-memory layer for teams whose code is increasingly written by coding agents that each start from zero. Its premise is that the expensive failure is not a bad edit but a cold agent: a session that has no idea what was decided last week, what is actively in flight, which parts of the repository a change touches, or what verification a similar change required. Snipara indexes that history — decisions, active work, impact and outcomes — and serves it as a Project Brain that every agent reads before its first edit, so context persists and compounds instead of being re-explained per session and per tool. It connects to repositories, docs and the tools a team already uses, keeps state current in real time as people and agents change things, and keeps an audit trail that traces why a decision was made and what happened after. A code graph with symbol cards and impact checks makes the scope of a proposed change explicit rather than inferred. Access is through a hosted MCP endpoint, a REST API, a GitHub App that syncs repositories and produces PR answer packs, and multi-agent primitives for coordinating swarms, claims, state and events. An optional local Companion CLI acts as a lead-engineer layer, turning project intelligence into scoped plans, routing guidance and proof requirements before an agent executes, and the server itself is published on GitHub under FSL-1.1-Apache-2.0 for teams that want to self-host.

Key Features

Tags