AI trading agent

Live

An AI agent wired into your signals and positions.

The agent reads the same signals, positions, and settings the bot runs on — so it can analyze a token, tell you why an auto-buy did or did not fire, and place a trade you confirm. It runs on your own OpenAI or Anthropic key.

Native tools

62

Internal MCP tools

13

LLM providers

OpenAI or Anthropic

Default execution

Confirmation required

Configuration

The guardrails, not just the capabilities

An agent that can move money needs hard limits. These are the ones ARCHR enforces regardless of how you phrase a request.

Trade cooldown

30s between agent trades

Auto-execute ceiling

25 trades/day, then confirmation

Trades per turn

1

Confirmation expiry

5 minutes

Sells

Always confirmed

Private keys

Never retrievable by the agent

Explainer

What is an AI trading agent?

Updated 2026-07-27

An AI trading agent is a conversational assistant that can actually reach your trading data and act on it, rather than just talk about markets in general. You ask a question in plain language, it calls tools against live data to answer, and — for anything that moves money — it hands you a confirmation instead of executing silently.

The distinction that matters is whether the agent is bolted on or built in. A general chatbot pointed at a price API can summarize a token. It cannot tell you what you hold, what your filters are set to, or why a trade you expected did not happen. ARCHR's agent reads the same position data, signal preferences, and trading plans the bot's own menus use, so asking it to change a setting changes that setting, and asking why an auto-buy did not fire lets it inspect the actual gates.

ARCHR's agent is also bring-your-own-key: you connect an OpenAI or Anthropic key and model usage bills to your account rather than through a subscription or credit system. You pick the provider and model tier, you see exactly what it costs, and you can revoke access by rotating one key. There is no incentive for us to quietly route you to a cheaper model than the one you asked for.

The part that matters most is the boundary around execution. Analysis is free to run; money movement is not. Sells always require an explicit confirmation button. Buys require one too unless you deliberately turn on auto-execute, and even then the agent falls back to confirmation after a daily threshold. An agent that is confidently wrong should cost you a declined confirmation, not a position.

The problem

Research does not fit on a keyboard

Evaluating a Solana memecoin means checking liquidity, holder concentration, whether the deployer has done this before, whether the token is even sellable, and where it sits in its lifecycle. That is five tabs and a few minutes, which is longer than most of these charts give you.

The usual bot answer is more buttons. That works until the number of things worth checking exceeds what fits on a keyboard layout, which happened a while ago.

How it works

Tools, not vibes

The agent has 62 native tools covering token data, portfolio state, signal configuration, wallet management, and trade requests, plus 13 internal MCP tools for on-chain and market context when those integrations are configured.

Each answer is assembled from tool calls against live data rather than from the model's memory. When an optional data source is unavailable the agent degrades gracefully and says so, instead of inventing a plausible number.

  • Token analysis with safety, holder, and lifecycle context
  • Position and P&L review in plain language
  • Signal preference and trading plan edits by chat
  • Trade requests that render as confirmation buttons

The limits

Where an AI agent is the wrong tool

Models miss fresh on-chain changes, over-weight whatever data they happened to retrieve, and sound equally confident whether or not they are right. None of that is fixed by a better prompt.

Treat the agent as a fast research assistant whose work you spot-check, and size positions as though its analysis is wrong. Keep confirmation on while you are learning how it behaves.

Common questions

Questions ARCHR answers

Direct answers, including the ones where the honest answer is “no”.

Does ARCHR pay for my AI usage?

No. ARCHR uses a BYOK model — you supply an OpenAI or Anthropic API key and pay that provider directly. ARCHR never bills LLM tokens.

Can the AI agent trade without asking me first?

Only if you explicitly enable auto-execute, and only for buys. Sells always require a confirmation button. Auto-execute also reverts to confirmation after 25 agent trades in a day.

Which AI models can I use with ARCHR?

Any model available on your OpenAI or Anthropic API key. Because you supply the key, you choose the provider and the model tier.

Is my API key safe?

Keys are AES-256 encrypted at rest and used only for your own agent sessions. The agent has no tool that can reveal a private key or an API key, and you can revoke access at any time by rotating the key with your provider.

What can the AI agent actually do besides answer questions?

It can analyze tokens, review positions and P&L, update signal preferences, edit trading plans, manage wallet settings, and request buys and sells that you confirm in Telegram.

How is this different from asking ChatGPT about a token?

A general chatbot has no access to your positions, your settings, or live Solana data, and cannot execute anything. The ARCHR agent reads the same live data the bot uses and can act on it after you confirm.

What it does

  • Analyze a token's holders, liquidity, deployer history, and safety flags in one message.
  • Review open positions, realized P&L, and portfolio state without opening a menu.
  • Change signal preferences, trading plans, and wallet settings by chat.

How traders use it

  • Check a token a friend just sent you before you size into it.
  • Ask why an auto-buy did or did not fire on a signal.
  • Rebalance exit settings across positions without clicking through submenus.

At a glance

Key model
Bring your own — you pay the provider directly
Key storage
AES-256 encrypted at rest
Providers
OpenAI, Anthropic
Execution default
Confirmation buttons in Telegram

An AI agent wired into your signals and positions.

ARCHR runs in Telegram. @archr_signal_bot — free to start, 1% platform fee on trades.

Memecoin trading is highly volatile and can result in total loss of capital. ARCHR provides software tools, not financial advice. Read the full risk disclosure and terms of use.