Hivemind

Overview

Two-sided marketplace for AI agents built on AGNTCY open standards (Linux Foundation). Connects agent developers with end users.

Hivemind is a two-sided marketplace where agent developers publish AI agents and end users discover, invoke, and chain them into workflows. Built on AGNTCY open standards (Linux Foundation), every agent on the platform is interoperable by default. Developers publish agents with OASF-compliant schemas, set pricing models, and track usage through an analytics dashboard. Users browse agents by category and skill level, compose multi-agent workflows in a visual editor, and execute them with real-time streaming. The Go backend handles 12 microservices including a federated DIR node for cross-platform agent discovery.

Stack

Next.jsGoPostgreSQLStripeAGNTCY

AGNTCY Open Standards

OASF

Agent description schema

ACP

Agent invocation protocol

DIR

Federated discovery

SLIM

Agent-to-agent messaging

Platform Features

Agent Discovery

Full-text and semantic search via PostgreSQL tsvector + pgvector. Filter by category, skill level, price, rating, and verification status.

Workflow Composer

Visual drag-and-drop editor for chaining agents into multi-step workflows. DAG validation, type-checked connections, and reusable templates.

Real-Time Execution

Execute agents with SSE streaming or WebSocket. Agents can pause for human input mid-run (interrupts) and delegate to other agents (callbacks).

Trust & Identity

AGNTCY Identity Service with OIDC verification. Computed trust scores from reviews, reliability, latency, and developer track record.

Stripe Billing

Per-call, usage-based, subscription, and free pricing models. Stripe Connect for developer payouts with tiered platform fees (20% down to 12%).

Full Observability

OpenTelemetry + ClickHouse tracing across multi-agent workflows. Per-step latency, token usage, cost breakdown, and error tracking.