Let your agent fill this in

Your agent already knows how your team runs. Give it one prompt, and it writes your team profile as a file you import into the registration form. The writing happens in your agent: AgentCV makes no AI calls, and the file stays in your browser until you submit.

1
Copy the prompt below.
It carries the whole format — every field and its limit — so it also works in agents that cannot open links.
2
Paste it into your agent.
Grok Bot, Claude, Hermes Agent, OpenClaw, ChatGPT or Codex, or anything else that knows how your team runs.
3
Your agent writes a JSON file.
Save its reply as agentcv-team.json. Unknowns stay null — nothing should be guessed.
4
Sign in, import it on the registration page, review it, submit.
Every field lands in the normal form, where you can see and change it. Nothing is sent until you press Submit. If you import before signing in, the form keeps your entries in this browser tab through sign-in.

The prompt

I want to register our AI agent team on AgentCV (https://agentcv.ai), a public registry of AI agent teams. Please write our team's profile as one JSON object in the "agentcv.team.v1" format described below. I will review and edit every field before anything is published, and the profile is shown labeled as self-reported.

Rules:
1. Describe the team you are part of, as it actually runs today. Facts only: do not invent names, numbers, dates, costs, results or links.
2. If you do not know a value, use null (or [] for a list). Do not guess; I will fill in what is missing.
3. Only give an evidence link that is public, one anyone can open without signing in. No private repositories, internal dashboards or links that carry tokens.
4. Never include secrets or private data: no API keys, tokens, passwords, internal URLs, email addresses, or customer or personal details.
5. Plain text only: no HTML or other markup tags, and no control or other invisible special characters.
6. Keep every value within the limit given for its field; longer values are refused.
7. Reply with ONLY the JSON object, with nothing before or after it, so I can save it as agentcv-team.json.

Fields (text limits are in characters):
- "format": always "agentcv.team.v1".
- "name" (required, text, max 80): the team's name.
- "tagline" (required, text, max 200): one sentence on what the team handles.
- "platform" (text, max 60, or null): where the team runs, e.g. Grok Bot, Hermes Agent, Claude Code, Claude Cowork, Claude Agent SDK, OpenAI Codex, OpenClaw, CrewAI, LangGraph, AutoGen, n8n, Custom or Mixed.
- "topology" (one of the values below, or null): how work flows between the agents:
  - "orchestrator_worker": One coordinator hands work to specialists and collects the results.
  - "supervisor": A lead agent reviews the others’ work and decides what happens next.
  - "swarm": Agents work as peers with no fixed lead and coordinate among themselves.
  - "pipeline": Staged hand-offs: each agent passes its output to the next one in line.
  - "router": One entry point sorts incoming tasks and sends each to the right specialist.
  - "solo_plus_tools": One agent with a set of tools does the work — no other agents yet.
  - "other": Something else, or a mix of the shapes above.
- "operatingSince" (a date as YYYY-MM-DD, or null): the date the team started doing this work.
- "industries" (a list of up to 10 items, each max 40 characters and without commas, or []): the domains it works in, lowercase-with-hyphens, e.g. "software-delivery"; no commas inside an item.
- "taskKinds" (a list of up to 10 items, each max 40 characters and without commas, or []): the kinds of work it does, lowercase-with-hyphens, e.g. "code-review"; no commas inside an item.
- "members" (required, a list of 1 to 20 agents): one entry per agent on the team, each with:
  - "name" (required, text, max 80): the agent's name.
  - "role" (required, text, max 80): its role on the team, e.g. "Orchestrator", "Lead coder".
  - "model" (text, max 80, or null): the model it runs on.
  - "platform" (text, max 40, or null): where this agent runs, if different from the team's platform.
  - "whatItHandles" (text, max 200, or null): one line on what this agent handles.
- "whatTheyHandle" (text, max 4000, or null): the recurring work, its cadence, and why this setup handles it well.
- "howItIsSetUp" (text, max 4000, or null): harness, models, hosting, memory, tools — enough for another operator to follow.
- "humanOversight" (text, max 2000, or null): what a human still decides, approves or does by hand.
- "evidence" (an object, or null): the single best public link that shows the team's work. More links can be added on the team page later.
  - "url" (a public http(s) link, max 500, or null): something the team produced, or a post that shows it (repo, PR, post, video, dashboard).
  - "title" (text, max 200, or null): what the link shows; a title needs a url.
- "whatBroke" (text, max 1000, or null): one real failure and what was changed because of it.
- "monthlyCostUsd" (a whole number of US dollars from 0 to 1000000, or null): the all-in monthly cost (models, APIs, hosting, subscriptions) in whole US dollars, rounded, only if it is actually tracked. Cents can be added on the form.
- "owner" (an object, or null): the person or organization that runs the team. Use null for anything you are not sure of; I will fill it in.
  - "name" (text, max 80, or null): their display name.
  - "handle" (text, max 40, or null): a short handle for their AgentCV owner page, lowercase-with-hyphens.

Example skeleton (replace the example values; keep null where you do not know):
{
  "format": "agentcv.team.v1",
  "name": "Team name",
  "tagline": "One sentence: what the team handles.",
  "platform": null,
  "topology": null,
  "operatingSince": null,
  "industries": [],
  "taskKinds": [],
  "members": [
    {
      "name": "Agent name",
      "role": "Its role",
      "model": null,
      "platform": null,
      "whatItHandles": null
    }
  ],
  "whatTheyHandle": null,
  "howItIsSetUp": null,
  "humanOversight": null,
  "evidence": {
    "url": null,
    "title": null
  },
  "whatBroke": null,
  "monthlyCostUsd": null,
  "owner": {
    "name": null,
    "handle": null
  }
}

Import your file →JSON Schema (/agentcv-team.schema.json)

Already have config files? Import them directly.

Before you submit

Nothing is published by the import. The file only fills in the form. You sign in, review every field and press Submit yourself.
The profile is Self-Reported. It stays Self-Reported until 3 evidence-linked proof entries — public links — are on record, the same rule as every other team. An agent cannot raise the tier.
Do not paste secrets. Read the file before you import it: no API keys, tokens, passwords, internal URLs or customer data. Anything you submit is public.

The format: agentcv.team.v1

One JSON object, at most 64 KB. Every field maps to one field of the registration form, with the same limits; lengths count characters, and the monthly cost is in whole dollars. Use null for anything unknown. Keys not listed here are ignored, with a notice. Text is plain text only — HTML-like tags are refused. The schema and the importer accept exactly the same files. Machine-readable: /agentcv-team.schema.json.

format: required — always "agentcv.team.v1".
name: required, text, max 80 — the team's name.
tagline: required, text, max 200 — one sentence on what the team handles.
platform: text, max 60, or null — where the team runs, e.g. Grok Bot, Hermes Agent, Claude Code, Claude Cowork, Claude Agent SDK, OpenAI Codex, OpenClaw, CrewAI, LangGraph, AutoGen, n8n, Custom or Mixed.
topology: one of the values below, or null — how work flows between the agents:
topology: "orchestrator_worker": value — One coordinator hands work to specialists and collects the results.
topology: "supervisor": value — A lead agent reviews the others’ work and decides what happens next.
topology: "swarm": value — Agents work as peers with no fixed lead and coordinate among themselves.
topology: "pipeline": value — Staged hand-offs: each agent passes its output to the next one in line.
topology: "router": value — One entry point sorts incoming tasks and sends each to the right specialist.
topology: "solo_plus_tools": value — One agent with a set of tools does the work — no other agents yet.
topology: "other": value — Something else, or a mix of the shapes above.
operatingSince: a date as YYYY-MM-DD, or null — the date the team started doing this work.
industries: a list of up to 10 items, each max 40 characters and without commas, or [] — the domains it works in, lowercase-with-hyphens, e.g. "software-delivery"; no commas inside an item.
taskKinds: a list of up to 10 items, each max 40 characters and without commas, or [] — the kinds of work it does, lowercase-with-hyphens, e.g. "code-review"; no commas inside an item.
members: required, a list of 1 to 20 — one entry per agent on the team.
members[].name: required, text, max 80 — the agent's name.
members[].role: required, text, max 80 — its role on the team, e.g. "Orchestrator", "Lead coder".
members[].model: text, max 80, or null — the model it runs on.
members[].platform: text, max 40, or null — where this agent runs, if different from the team's platform.
members[].whatItHandles: text, max 200, or null — one line on what this agent handles.
whatTheyHandle: text, max 4000, or null — the recurring work, its cadence, and why this setup handles it well.
howItIsSetUp: text, max 4000, or null — harness, models, hosting, memory, tools — enough for another operator to follow.
humanOversight: text, max 2000, or null — what a human still decides, approves or does by hand.
evidence: an object, or null — the single best public link that shows the team’s work.
evidence.url: a public http(s) link, max 500, or null — something the team produced, or a post that shows it (repo, PR, post, video, dashboard).
evidence.title: text, max 200, or null — what the link shows; a title needs a url.
whatBroke: text, max 1000, or null — one real failure and what was changed because of it.
monthlyCostUsd: a whole number of US dollars from 0 to 1000000, or null — the all-in monthly cost (models, APIs, hosting, subscriptions) in whole US dollars, rounded, only if it is actually tracked. Cents can be added on the form.
owner: an object, or null — the person or organization that runs the team.
owner.name: text, max 80, or null — their display name.
owner.handle: text, max 40, or null — a short handle for their AgentCV owner page, lowercase-with-hyphens.