Local Agentic Query Surface

❯ Queries are assets.
Treat them that way.

The first query surface where humans AND agents share execution, approvals, and output artifacts.

Anonymous browser mode. Free forever. No account. No install. Just open it.

Open Spürhund See Pricing

spuerhund — agentic query surface

❯ spuerhund run gql --server prod --saved-query order-status

Session sess_01HS7Y8XQ7A6C2

Observer approved in Spürhund UI

Output json · 2.8 MB

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

Warning inline output truncated, full result kept local

Server https://api.store.com/graphql

Auth Bearer token resolved from vault

Response 200 OK (89ms)

"preview" :

"ordersReturned" : 8421 , "firstStatus" : "shipped"

❯

The Problem

Your workflow is held together with copy-paste.

Queries scattered across tabs, Slack threads, CI scripts, and forgotten READMEs. Auth setups that vanish with your browser session. Agents that burn tokens rediscovering what you already know.

Queries live in five places

GraphiQL tab. Postman collection. Curl in your terminal history. That Slack message from three weeks ago. None of them talk to each other.

Auth dies with your tab

You spent 20 minutes configuring OAuth. Then you closed the browser. Gone. Tomorrow you'll do it again. And again.

Agents rebuild everything

Your AI agent discovers the schema, figures out auth, constructs the query from scratch. Every. Single. Time. Tokens wasted on solved problems.

CORS blocks real work

The browser sandbox stops you from reaching local services, VPN targets, and internal APIs. So you switch to curl. Again.

The Answer

Interactive, terminal, and agent surfaces. One workspace.

Same servers. Same saved queries. Same vault-backed config. Browser first, Local Runtime underneath, with observer-aware CLI and MCP execution on top.

Browser App

Full-featured workspace in the browser. Manage servers, saved queries, schema data, and vault-backed config in one place.

GraphQL + REST in one interface
Schema browser and runtime-backed catalogs
Encrypted secret vault
Dynamic auth with scripts
Saved, named, categorized queries

CLI

Your runtime workspace from the terminal. Session-aware execution, observer approvals, and JSON output that stays model-safe when results explode.

Run saved queries with session IDs for agent threads
Search schemas from terminal
Always writes local scratch artifacts for execution output
JSON output with size, type, and scratch-file metadata
Create and update servers and queries locally
Same workspace, different surface

MCP Bridge

AI agents get structured access to the runtime workspace. Discover servers, inspect saved queries, execute known-good paths, and route execution through local observers.

Agents use what you already built
Structured tool discovery
Observer-aware approvals for agent execution
Large outputs handed off as local scratch artifacts
No token waste on rediscovery
Works with any MCP client

❯ Queries are assets.Treat them that way.