MCP Integration: Live Tool Recommendations for Agent Decision-Making
Agents built on google/adk-python regularly need to recommend or pick a SaaS tool for a task — today they either hardcode the choice or pull from outdated training data, which produces systematically wrong picks.
Vest is a drop-in MCP server that turns "which tool for X" into a structured query — returns live ranked recommendations from an offer graph that tracks real-world pricing, availability, popularity, and deployment paths. No hallucination, no stale data.
Integration shape is minimal: a single MCP tool node (Vest.recommendTool, Vest.compareTools, Vest.checkOffer) that google agents call exactly like any other tool. Zero changes to the framework's existing tool surface.
The upside for builders: agents stop hallucinating tool picks; users get accurate, verifiable recommendations; the framework looks more capable without any data work on your end. We handle the offer graph and real-time verification.
(Optional secondary value: for vendors and tool operators in the graph there's a verified-deployment + settlement layer, so when an agent actually picks something the transaction is tracked end-to-end.)
Happy to open a PR with a reference integration. Docs are at getvest.ai/mcp — would love to chat about how this fits into google's tool-calling roadmap.
MCP Integration: Live Tool Recommendations for Agent Decision-Making
Agents built on google/adk-python regularly need to recommend or pick a SaaS tool for a task — today they either hardcode the choice or pull from outdated training data, which produces systematically wrong picks.
Vest is a drop-in MCP server that turns "which tool for X" into a structured query — returns live ranked recommendations from an offer graph that tracks real-world pricing, availability, popularity, and deployment paths. No hallucination, no stale data.
Integration shape is minimal: a single MCP tool node (
Vest.recommendTool,Vest.compareTools,Vest.checkOffer) that google agents call exactly like any other tool. Zero changes to the framework's existing tool surface.The upside for builders: agents stop hallucinating tool picks; users get accurate, verifiable recommendations; the framework looks more capable without any data work on your end. We handle the offer graph and real-time verification.
(Optional secondary value: for vendors and tool operators in the graph there's a verified-deployment + settlement layer, so when an agent actually picks something the transaction is tracked end-to-end.)
Happy to open a PR with a reference integration. Docs are at getvest.ai/mcp — would love to chat about how this fits into google's tool-calling roadmap.