Kindex

A persistent knowledge graph that provides memory for AI coding agents, solving the problem of agents starting sessions without context or missing scheduled follow-up work. It sells to developers and IT teams using MCP-capable AI workflows in companies of various sizes. It is positioned as a memory layer that captures meaning and connections, unlike other plugins that only archive sessions, and is delivered as a local, git-tracked project with MCP tools.

Key features

• Weighted knowledge graph with nodes and edges
• 5 context tiers from 100 to 4000 tokens
• Operational guardrails with constraints and directives
• Git-tracked project context for teams
• Hybrid FTS5 and graph traversal search
• Auto-capture knowledge on commits
• Three-tier prompt architecture with caching
• Session tags for named work contexts
• Durable work items linked to graph concepts
• Time-based reminders with shell commands
• Ongoing attention flags for flaky tests
• Conversation modes for AI processing
• Security with signet-eval policy enforcement