Publish your research as an AI agent.

Before you start

Your submission must meet these conditions. Agents that don't meet them will be rejected at validation time.

• 1.A published, peer-reviewed paper with a live DOI. Preprints are not accepted. The DOI must resolve to a publicly accessible abstract or full text. The agent is validated against this DOI on submission.
• 2.A public GitHub repository. Your code must be publicly accessible. Cited pins an exact commit hash at publish time — nothing is copied, only referenced.
• 3.An entrypoint that reads JSON from stdin and writes JSON to stdout. The agent must accept input as a JSON object piped to stdin and return a JSON object on stdout. Any language is supported.
• 4.A dependency file. Include a requirements.txt, pyproject.toml, package.json, or equivalent so the platform can reproduce your environment.
• 5.An open-source license. Include a LICENSE file in your repo. MIT, Apache 2.0, and GPL are all accepted.

Step 1 — Prepare your repository

Your repo should contain your research implementation with a clear entrypoint script. The script must accept a JSON object on stdin and return a JSON object on stdout. A minimal Python example:

import json, sys

data = json.load(sys.stdin)

# your logic here
result = {"output": "..."}

print(json.dumps(result))

Any language is supported as long as stdin/stdout JSON contract is met.

Step 2 — Add an agent.yaml

Place an agent.yaml file in the root of your repository. This is the manifest that Cited reads to validate and index your agent.

name: your-agent-slug           # lowercase, hyphens only
title: "Human-readable title"
description: "One or two sentences describing what this agent does."
version: "1.0.0"

doi: "10.xxxx/your-doi-here"   # must be a live, peer-reviewed DOI
entrypoint: "python src/main.py"

license: MIT                   # SPDX identifier

input_schema:
  type: object
  properties:
    query:
      type: string
      description: "Your input field"
  required: [query]

output_schema:
  type: object
  properties:
    result:
      type: string
  required: [result]

examples:
  - input:
      query: "example input"
    description: "A representative example"

sdk: anthropic                 # anthropic | openai | google | local | other

Step 3 — Submit via the dashboard

Once your repo has an agent.yaml and meets the requirements above, register and submit from your dashboard.

1. Create an account or log in at citedagents.com.
2. Go to your dashboard and click Publish Agent.
3. Enter your GitHub repo URL and the DOI of your paper.
4. The platform clones your repo, resolves the DOI, validates the manifest, and indexes the agent automatically.
5. Once live, your agent appears in the catalog with a permanent DOI attribution link.

What happens at validation

Cited runs four checks before your agent goes live:

• ✓ DOI resolved — the DOI is fetched and must return metadata from a recognised publisher.
• ✓ Manifest valid — the agent.yaml passes schema validation. All required fields must be present.
• ✓ Schema verified — input_schema and output_schema are valid JSON Schema objects.
• ✓ Indexed & live — the agent is pinned at the current commit hash, assigned a slug, and published in the catalog.

If validation fails, you will receive a specific error message explaining which check failed and how to fix it.

Academic Program — Free Pro access

Researchers who successfully publish a verified agent receive full Pro access free. This includes hosted execution quota, a researcher badge, and DOI attribution on every run. No credit card required.

Pro access is granted automatically after your first agent passes validation.