Kaito Query Service
Route agent tasks through one MCP gateway to Gemini, MiniMax, Replicate, and OpenRouter for vision, search, and code review.
Overview
Kaito Query is a MCP server for the Build phase that connects agents to multiple LLM and tool providers for vision, search, and code review via hosted HTTP transport.
What is this MCP server?
- Remote streamable-http MCP at us-central1-sand-gallery-lab.cloudfunctions.net (manifest lists 2 remote URLs)
- Server version 1.2.0 titled Kaito Query Service
- Supports Gemini, MiniMax, Replicate, and OpenRouter-backed workflows from one surface
- Advertised capabilities include vision, search, and code review for coding agents
- README notes USDC on Base for paid or metered access patterns
- Server version 1.2.0 in MCP manifest
- 2 streamable-http remote URLs on Google Cloud Functions
- 4 named provider families in description: Gemini, MiniMax, Replicate, OpenRouter
What problem does it solve?
Juggling separate API keys and SDKs for Gemini, Replicate, OpenRouter, and vision models slows agent setup and fragments how you pay and monitor usage.
Who is it for?
Builders who want a single remote MCP bridge to several model vendors and agent-friendly review/search tools with Base USDC billing.
Skip if: Teams with strict data residency that require fully offline or self-hosted inference only, or builders avoiding crypto-settled services.
What do I get? / Deliverables
After you add the Kaito Query remote MCP URLs, your agent can call unified tools for models, vision, search, and review without bespoke integration code in your repo.
- Two registered remote MCP endpoints (mcp and toolsCall URLs)
- Agent-accessible unified LLM, vision, search, and review tools
- Reduced custom SDK code for Gemini, MiniMax, Replicate, and OpenRouter
Recommended MCP Servers
Journey fit
Kaito Query sits in Build integrations because builders connect it while wiring multi-provider AI capabilities into their agent stack. Integrations is the right subphase for a remote MCP that abstracts several third-party model and tool APIs behind one endpoint.
How it compares
Hosted multi-provider LLM gateway MCP, not a local Ollama skill or a single-vendor Claude-only plugin.
Common Questions / FAQ
Who is kaito-query for?
Kaito Query is for solo builders and agent users who want one MCP entry point to Gemini, MiniMax, Replicate, OpenRouter, vision, search, and code review.
When should I use kaito-query?
Use it when you are integrating diverse model APIs into an agent workflow and prefer remote MCP tools over maintaining multiple client libraries yourself.
How do I add kaito-query to my agent?
Add the published streamable-http remote URLs from the server manifest to your MCP client, verify connectivity to the Sand Gallery cloud function endpoints, and fund USDC on Base if your usage requires it.