Connects Claude Desktop to BrainstormRouter's governed AI routing layer. Exposes multi-provider chat completions across 36 models from Claude, GPT, Gemini and five other providers, with Thompson sampling that learns which models perform best for your workload. Every request returns counterfactual cost receipts showing what static routing would have spent. You get per-agent budget caps, semantic caching via pgvector, and memory injection across six context attributes (SOUL, MEMORY, SKILL, etc.). Useful when you want intelligent model selection without manual fallback chains, or when you need cost governance and audit trails baked into the routing layer instead of bolted on after.
claude mcp add --transport stdio justinjilg-brainstormrouter uvx brainstormrouter