Contextstream Memory Code Search

ContextStream Persistent Memory & Code Search Skill (contextstream)

ContextStream provides AI coding assistants with long-term cross-session memory, semantic codebase search, dependency graph analysis, pre-compaction state checkpointing, and team knowledge integration via https://mcp.contextstream.io/mcp.

Domain Tool Surface

1. search: Hybrid semantic + keyword code search with exact-token symbol fusion.
2. memory: Persists decisions, lessons, conventions, runbooks, and project plans across sessions.
3. session: Captures critical state before context window compaction; lists and recalls past session transcripts.
4. qa: Grounded Q&A over project memory and conventions with citation receipts.
5. graph: Codebase dependency mapping, impact analysis, dead code, and circular dependency detection.
6. capsule: Packages project state into portable, shareable context snapshots for agent handoffs.

Best Practices for AI Agents

1. Pre-Compaction Capture: Automatically trigger session state capture before context window compaction to prevent losing design intent and architecture decisions.
2. Semantic Search: Use search for intent-based queries ("where do we handle auth") instead of brute-force file scanning.
3. Optional CONTEXTSTREAM_API_KEY for index scale, graph features, and team integrations.