Builder Docs— Documentation
Using AgentsB2B DocsAll Docs →

Overview

As a builder, you publish AI agents to the FindUsefulAgents marketplace. Consumers run your agents and pay with credits. You earn 70% of every run — the platform takes 30% to cover infrastructure, payments, and discovery.

🤖

Publish agents

Submit any AI agent — from a simple Bedrock call to a multi-step Strands workflow.

💰

Earn 70%

Every paid run credits your builder account. Payouts are processed monthly.

📊

Track earnings

Builder portal shows per-agent earnings, run counts, and payout history.

Getting Started

  1. 1

    Activate your builder account

    Sign into your consumer account and click "Build" in the top navbar. Your builder account is created automatically — no separate registration needed.

  2. 2

    Open the Builder Portal

    The portal at /builder/portal gives you a dashboard, agent management, earnings tracker, and payout settings.

  3. 3

    Create your first agent

    Go to My Agents → New Agent. Choose an agent type, configure the model/prompt/endpoint, set a credit price, and submit for review.

  4. 4

    Agent review

    All new agents go through a brief review (usually < 24h) before appearing in the marketplace. You'll receive an email on approval or rejection.

  5. 5

    Start earning

    Once live, consumers can discover and run your agent. Earnings appear in your dashboard in real-time.

Agent Types

Choose the right agent type for your use case. All types run on our infrastructure — you just configure behaviour.

Third-Party (HTTPS Endpoint)

You host the AI logic on your own server. We send the user's inputs as a POST request and display your JSON response. Full control — you can use any model, API, or processing pipeline.

POST https://your-server.com/your-agent
Content-Type: application/json

{ "input": { ...user_inputs } }

# Expected response:
{ "output": "...", "output_type": "text" }

Bedrock (Text / Image)

Configure a prompt template and we run it on Amazon Bedrock. Great for text transformation, Q&A, classification, and image generation. No server needed.

Agent config:
  model_id: us.amazon.nova-pro-v1:0
  system_prompt: "You are a competitive analysis expert..."
  prompt_template: "Analyse {{company}} in the {{industry}} sector."

Strands (Multi-Step Workflow)

Build agentic workflows with tool use (web search, calculators, APIs). The model reasons, calls tools, and iterates to produce a result. Best for research and complex tasks.

Agent config:
  model_id: us.amazon.nova-pro-v1:0
  tools_enabled: ["web_search", "calculator"]
  system_prompt: "Research the company and provide a full report."
  max_loop_count: 5

Chatbot (Stateful)

Multi-turn conversation agent. Session history is preserved across messages. Ideal for assistants, support bots, and guided workflows.

Agent config:
  b2b_agent_type: chatbot
  system_prompt: "You are a helpful customer support agent for..."

# Each call includes last 50 turns of history automatically

Image Generation

Text-to-image or image-to-image agents powered by Amazon Titan. The user provides a prompt (and optional reference image) and gets a generated image back.

Agent config:
  problem_type: text_to_image
  model_id: amazon.titan-image-generator-v2:0
  prompt_template: "Generate a professional logo for {{company_name}}: {{description}}"

Creating an Agent

Go to My Agents → New Agent in the builder portal. Fill in the following fields:

Name

required

Public name shown in the marketplace. Keep it concise and descriptive.

Description

required

What the agent does and what it outputs. Used in search and on the agent detail page.

Agent Type

required

Choose: third_party, text_to_text, text_to_image, strands_agent.

Endpoint URL

optional

Required for third-party agents. Must be HTTPS and respond within 60 seconds.

Model ID

optional

For Bedrock/Strands agents. Defaults to Nova Lite. Use Nova Pro or Claude Sonnet for complex tasks.

System Prompt

optional

Instructions for the AI. Defines persona, output format, and constraints.

Prompt Template

optional

Template with {{variable}} placeholders that map to your input schema fields.

Input Schema

required

JSON Schema defining the fields the user fills in before running the agent.

Credit Cost

required

Credits charged per successful run. Most agents charge 1–5 credits. Higher for complex/expensive models.

Thumbnail

optional

Square image (512×512+) shown in the marketplace. Agents with thumbnails get more clicks.

Input schema uses standard JSON Schema format. Each field becomes a form input the consumer fills in before running.
// Example input schema for a competitor research agent
{
  "type": "object",
  "properties": {
    "company_name": { "type": "string", "title": "Company Name" },
    "focus_area": {
      "type": "string",
      "title": "Focus Area",
      "enum": ["pricing", "funding", "tech stack", "team"],
      "default": "pricing"
    }
  },
  "required": ["company_name"]
}

Earnings & Payouts

Every paid run of your agent creates a transaction. You earn 70% of the gross credit value, converted to USD at $0.01/credit.

Credit value

$0.01 / credit

Your share

70%

Platform fee

30%

Minimum payout

$25

# Example: agent priced at 3 credits
Consumer pays:  3 credits = $0.03
You earn:       $0.03 × 70% = $0.021 per run

# At 1,000 runs/month:
Monthly gross:  1,000 × $0.03 = $30
Your earnings:  $30 × 70% = $21

Payout Setup

Go to Settings in the builder portal and enter your payout method. Supported methods:

PayPal — Your PayPal email address. Most common.
Bank Transfer (ACH/Wire) — Account number + routing number. US accounts only.
USDC (Crypto) — ERC-20 or Solana wallet address. Near-instant settlement.
Payouts are processed monthly. Minimum threshold is $25. Earnings below $25 roll over to the next month.

Managing Agents

Agent statuses

draft → pending_review → live (or rejected). Edit and resubmit rejected agents. Live agents can be paused from the portal.

Editing a live agent

Minor edits (description, thumbnail, credit price) take effect immediately. Changes to prompts or endpoints go back to pending_review.

Viewing per-agent earnings

My Agents → click any agent → shows run count, total earned, and a timeline of recent transactions.

Suspending an agent

You can pause/suspend your own agents anytime from My Agents. Suspended agents don't appear in the marketplace.

Agent rejection

Rejections include a reason. Common reasons: unsafe outputs, misleading description, endpoint unreachable, or NSFW content.

FAQ

Do I need to host anything?

Only for third-party agents. For Bedrock and Strands agents, we run everything on our infrastructure. No server required.

Can I set my own pricing?

Yes. You choose the credit cost per run. Higher-quality or expensive-to-run agents can charge more.

How does the review process work?

We manually review each new agent for quality, safety, and accurate description. Approval typically takes < 24 hours.

Can I have multiple agents?

Yes, unlimited agents per builder account.

What models are available?

Amazon Nova Lite, Nova Pro, Claude Sonnet (via Bedrock), Amazon Titan for images, and any model your own endpoint supports.

Do I need a separate builder account?

No. Any consumer account can activate builder access instantly from the navbar. Click "Build" and your builder account is created automatically.

Share feedback

How would you rate your experience?