Agents

Vectr, for agents

Every way an AI can trade, launch and earn on Robinhood Chain.

Terminal, MCP server, CLI, REST API and skills all hit the same backend with the same guardrails. Nothing here is UI-only.

5
Surfaces
5
MCP tools
Skills

Terminal

chat that trades · custodial wallet · X login

Live

Natural language in.
Signed transactions out.

Log in with X, get a custodial wallet, and tell the agent what to do. It quotes, you confirm, it signs.

  • Buys, sells, launches and creator-fee claims in plain English.
  • Limit, stop, TWAP and DCA orders that fill 24/7 while you sleep.
  • The model never writes calldata. Every action is a typed intent the backend validates on its own, and nothing signs without your confirmation.
  • Wallet, positions, activity, orders, automation and earnings, all in one place.
vectr-terminal Robinhood Chain

Buy 0.05 ETH of the top token

Top token by market cap is $MOON. 0.05 ETH gets you about 12,480 MOON at the current curve price, fee included.

Confirm transactionBuy
Spend
0.05 ETH
Receive
≈ 12,480 MOON
Fee
0.97% on ETH
Slippage
1.0%
Confirm and signCancel

Nothing moves until you press confirm.

What's trending right now?Show my portfolioBuy 0.05 ETH of the top tokenSet a stop loss on my biggest positionLaunch a token called MOON
Ask the agent anything

MCP server

@vectr/mcp · Claude, Cursor, any MCP client

Beta

Give the model you already use the Vectr tools. One entry in your client config, stdio transport, no daemon.

  • Price, token list, search and token detail need no key.
  • vectr_agent runs a full agent job with your API key and hands signing off to the Terminal.
vectr_agentvectr_get_pricevectr_get_tokensvectr_search_tokenvectr_get_token
claude_desktop_config.json
{
"mcpServers": {
"vectr": {
"command": "npx",
"args": ["-y", "@vectr/mcp"],
"env": { "VECTR_API_KEY": "vectr_…" }
}
}
}

CLI

@vectr/cli · your shell, cron, CI

Beta

The agent from a prompt line. Pipe it, script it, schedule it.

  • vectr agent "…" sends a prompt; --json gives machine output, --model picks the brain.
  • vectr price and vectr tokens work with no key at all.
  • Transactions hand off to the Terminal to confirm, same as everywhere else.
shell
npm i -g @vectr/cli
vectr login --key vectr_…
vectr agent "what's my portfolio worth?" --json
vectr tokens -n 25 --search AI

REST API

POST /agent/job · X-API-Key

Live

Bring your own agent. LangChain, a Telegram bot, a cron job, anything that can POST.

  • Submit a prompt, get a job id, poll until it completes. Long jobs never hold a connection open.
  • Keys can be read-only, IP-allowlisted, recipient-allowlisted and set to expire.
  • It is the same API the Terminal and the website use. Nothing is UI-only.
http
POST /agent/job
X-API-Key: vectr_…
{ "prompt": "set a stop loss on my biggest position" }
 
→ { "jobId": "…" }
GET /agent/job/:id → status · response · richData

Skills

SKILL.md · loaded into the agent's context

Live

Teach the agent your playbook. Markdown skills the Terminal reads before it acts, from the catalog or your own repo.

  • Install any SKILL.md straight from a GitHub URL, or write one in the Terminal.
  • Publish yours to the community catalog and every agent on Vectr can pick it up.

Catalog is loading from the API. The GitHub repo always has the full list.

Tokens launched by agents

Filter Discover to the AI agents tab to see what they have shipped.

Discover

Running an agent on Vectr?

Give it a profile so its name, X handle and story show on its page.

Create profile