How It Works — Spürhund

Product Walkthrough

One workspace for every API path.

Connect servers, organize queries, and explore schema or catalog data. HTTP works in the browser or runtime. MongoDB and PostgreSQL run through the Local Runtime.

Add a Server

Connect HTTP, MongoDB, or PostgreSQL targets. Store auth and connection config once. Reference vault secrets everywhere.

Organize Queries

Save, name, and categorize queries by project or team. Pin the ones you use daily. Add remarks and risk levels. Find anything instantly.

Run a Query

Run GraphQL, REST, Mongo, or SQL work with saved variables, headers, remarks, and risk levels. Open multiple queries side by side.

Explore Schemas

Browse GraphQL schemas and runtime-backed database catalogs. Inspect types, collections, tables, fields, and descriptions before you query.

GraphQL and REST can run directly in the browser. MongoDB and PostgreSQL execute through the Local Runtime.

Terminal Native

Same workspace. Different surface.

The CLI reads the local runtime workspace snapshot. Every execution writes a local scratch artifact, and large results come back with size, type, and file metadata instead of blowing up your model context.

spuerhund CLI

$ spuerhund list servers

Mode: signed-in

prod-api [prod-api] · 12 saved queries · https://api.example.com

analytics [analytics] · 4 saved queries · public (postgres)

$ spuerhund run gql --server prod-api --saved-query order-status

Status: 200

Duration: 142ms

Output: json · 2.8 MB

Scratch: ~/.spuerhund/scratch/outputs/2026-04-14/gql-sess_01HS7Y8XQ7A6C2.json

Warning: inline output truncated, full result kept local

{"preview":{"order":"ORD-42871","status":"shipped"}}

$ spuerhund schema search shipment --server prod-api

Shipment type 12 fields

ShipmentStatus enum PENDING | SHIPPED | DELIVERED

ShipmentInput input 6 fields

Agent-Native

Your agents are wasting tokens.

Every time an AI agent needs to query an API, it rediscovers schema, auth, and query shape. Spürhund gives agents a structured execution surface with session threading, observer approvals, and scratch artifacts for oversized results.

✖ Without Spürhund

1Agent introspects full schema
2Discovers auth requirements
3Builds token acquisition flow
4Constructs query from scratch
5Hopes it works. Retries.
6Repeats everything next session

✔ With Spürhund

1Agent calls spuerhund_list_queries
2Picks existing saved query and keeps the session ID
3Runtime notifies a connected observer UI before execution
4Observer approves or rejects the follow-up query
5Large outputs land in a local scratch file, not in the model context
✓Done, with traceable execution.

MCP Bridge -- Agent Interaction

// Agent discovers available saved queries via MCP
const queries = await mcp.call("spuerhund_list_queries", {
  server: "prod-api"
});

// Agent executes an existing saved query
const result = await mcp.call("spuerhund_execute_saved_query", {
  server: "prod-api",
  query: "order-status",
  sessionId: "sess_01HS7Y8XQ7A6C2",
  variables: { orderId: "ORD-42871" }
});

// → { status: 200, outputType: "json", size: "2.8 MB", scratchFile: "~/.spuerhund/.../sess_01HS7Y8XQ7A6C2.json" }