Ctx

ctx addresses a gap most agent users have felt but few tools solve: recovering past context. When an agent starts a new session, it has no memory of previous decisions, failed attempts, or commands. ctx indexes those logs and makes them searchable, so the next agent can pick up where the last one left off. Its local-first approach is a strong point — no cloud, no API key, no background service. The 50× token efficiency claim over raw transcript search is plausible given the structured indexing. It supports a growing list of agent harnesses including Claude Code, Codex, Cursor, Copilot CLI, and newer ones like Pi, OpenCode, and Factory. Where ctx falls short is discoverability: users need to explicitly install the skill for agent integration (npx skills add ctxrs/ctx). The free tier is strictly local; shared team history requires a cloud tier that isn't detailed on the site. If you only use one agent and rarely revisit sessions, ctx may be overkill. Its real value is for power users juggling multiple agents or complex, multi-session projects. Compared to agent memory tools like Mem0 or Graphify-style codebase graphs, ctx is more transparent — it retrieves raw transcripts rather than compressed summaries. For debugging why an agent chose a certain path, that's invaluable. Caveat: the tool is young, so provider support may lag behind new agent releases. But the MIT license means anyone can extend it.