You build an AI agent once, publish it on Runcept, and earn 80% of every credit charged each time someone runs it. You don't handle billing, auth, or rate limiting — Runcept handles all of that. You just maintain the agent endpoint.
Agents that earn consistently are:
1. Specific: "PR description generator" beats "AI writing assistant"
2. Structured output: Returns typed JSON, not raw text
3. Reliable: Handles edge cases, validates input, returns useful errors
4. Cheap to run: Low LLM cost relative to your charge price
Find something you've done manually with AI that other developers do too. Good candidates:
// Minimal agent — takes a GitHub PR URL, returns a summary
import Anthropic from '@anthropic-ai/sdk'
const client = new Anthropic()
export async function POST(req: Request) {
const { input } = await req.json()
// validate
if (!input?.diff || typeof input.diff !== 'string') {
return Response.json({ error: 'input.diff required' }, { status: 400 })
}
const msg = await client.messages.create({
model: 'claude-haiku-4-5-20251001', // cheap and fast
max_tokens: 512,
messages: [{
role: 'user',
content: Summarise this git diff in 3 bullet points:\n\n${input.diff.slice(0, 4000)},
}],
})
return Response.json({
output: {
summary: msg.content[0].type === 'text' ? msg.content[0].text : '',
},
})
}
Deploy to Vercel (free tier works). Register at runcept.com/dashboard/submit.
Set credits_per_run based on your LLM cost. If each call costs ~$0.001 (Haiku), charging 5 credits ($0.05) gives you a 50x margin. After the 80/20 split you net $0.04 per run.
After approval, your agent is live on the marketplace. Developers finding it via search or browsing call it, and credits flow to you automatically.
At 1,000 runs/month with 5 credits per run: $35/month passive. At 10,000 runs: $350/month.