Set Up OpenClaw to Trade on Kraken (Kraken CLI + MCP)

In March 2026, Kraken did something no other major exchange had bothered to ship: an official command-line interface with roughly 134 commands and a built-in MCP server. That last part is the one that matters for anyone building with agents. An AI agent can now read Kraken's markets and place trades through natural-language tools, on one of the oldest and largest centralized exchanges in crypto. This guide walks the whole setup: sign up for OpenClaw Direct, mint a Kraken API key scoped to trade but not withdraw, wire the Kraken CLI/MCP into a hosted Instance, bolt on a safety skill, and start read-only with Telegram approvals before a single order fills.

Why an Always-On Market Needs a Hosted Agent

Crypto never closes. Kraken's order books reprice every second of every day, and a setup you liked at midnight can be a different market by the time your alarm goes off. That's the whole problem a trading agent is supposed to solve, and it's also the problem a laptop-bound bot quietly fails at. A script that only runs when your machine is awake is a script that sleeps through the move that mattered.

So where should the agent actually live? Not on your laptop, and not in a terminal tab you'll close by accident. OpenClaw Direct runs every agent on its own dedicated machine, isolated from every other user, with your Kraken API key encrypted at rest inside the platform's credential vault instead of sitting in a plaintext config next to your browser. The infrastructure holds 99.9% uptime with 24/7 monitoring, so the agent keeps watching a market that also never rests, and you can suspend or terminate it from any browser on any device. For software that signs trades with real money, that's the right amount of surface area and the right size of blast radius.

The wedge is simple. A 24/7 market wants a 24/7 operator, and no human is that. A hosted Instance is.

What Is the Kraken CLI, and How Do Agents Plug Into It?

Kraken is a large, long-established centralized exchange — spot and derivatives, years of operating history, and one of the more recognizable names in the space. It also runs xStocks, a line of tokenized equities that puts stock-like exposure onchain. What changed in early 2026 is how an agent connects to all of it. The Kraken CLI ships with roughly 134 commands and a built-in MCP server, which is the part that turns a command-line tool into an agent-native one.

Here's the distinction worth internalizing. A CLI is something a human types into a terminal. An MCP server is something an AI agent talks to — a standard handshake that exposes a menu of tools the agent can call in plain language. Kraken bundling both means you don't have to build a translation layer: the same tool that lets you type kraken commands also lets an OpenClaw agent read the market and, when you allow it, act on it. For the full picture of how that handshake works, see our MCP explainer.

Because the MCP is model-agnostic, you're not locked to one AI provider. The agent connects, sees the Kraken tools it's permitted to call, and — crucially — shouldn't call any of the trading ones yet. The read side comes first.

What Can Your Agent Actually Do — and What Should It Not?

Through the Kraken CLI's MCP, an agent can read the full market and, with a trade-scoped key, act on it. The read side is close to free and low-risk; the trade side is where a careless agent turns a bad idea into a filled order. Respect the gradient between them.

• Read and analyze. Pull prices, order books, balances, and open orders, then write a brief on what it sees. Start here, and stay a while.
• Small spot orders. A modest position in a blue-chip pair is the right first live action once the reads look sane.
• Everything larger. Bigger size, more pairs, tighter loops — earn your way there with evidence, not enthusiasm. The agent should not be discovering new markets to trade on its own while you're still learning its behavior.

The safety skill in a later step exists to keep your agent on the left end of that gradient until you've decided, with evidence, to let it move right. Capability is not judgment, and the two arrive separately.

The Setup, Step by Step

Step One: Sign Up for OpenClaw Direct and Provision Your Instance

Go to openclaw.direct/users/sign_up and create an account. The Advanced plan ($29/month, "Most Popular") is the right starting point for trading workloads — 4 GB of dedicated RAM and unlimited AI models, so you're not rationing calls while your agent watches an around-the-clock market. Every subscription includes free trial credits, enough to wire the whole thing up and watch a few actions before you commit. Create a new Instance, name it something like "Kraken Agent," and start provisioning. While it builds, move to Step Two.

Step Two: Mint a Kraken API Key — Trade-Only, No Withdrawals

This is the step that decides your blast radius, so slow down here. In your Kraken account settings, create a new API key and scope it deliberately. Grant the permissions the agent needs to read markets and place or cancel orders. Do not grant withdrawal permission. A key that can trade but cannot move money off the exchange is a key that, even fully compromised, can only trade badly — it can't send your balance to an attacker's address. That single toggle is the most important safety decision in this entire guide.

This is standard exchange custody: your funds stay at Kraken, under your account, exactly as they would if no agent existed. The API key is a permissioned signer, not a handover of your coins. Once minted, the key and secret go straight into the OpenClaw Direct credential vault, scoped to your Instance and never written to a personal disk. If you ever find yourself pasting an API secret into a plaintext file, stop — that's the failure mode the vault exists to prevent.

Step Three: Wire the Kraken CLI/MCP Into Your Instance

With your Instance provisioned, add the Kraken MCP to its server list and point your agent at it. Supply the trade-scoped API key from Step Two from the vault, or start in read-only mode so the agent can see markets without any ability to trade. Because Kraken bundles the MCP into the CLI itself, you're wiring one well-supported interface rather than stitching together a third-party server. Once connected, your agent can see the Kraken tools it's allowed to call. It just shouldn't call the trading ones yet.

If you're weighing this against other venues, we keep a running comparison of the best MCP servers for traders, and a companion walkthrough for the onchain perps crowd on setting up an agent on Hyperliquid. Kraken's pitch is different: it's the established, custodied, spot-and-derivatives exchange with a first-party CLI, which makes it a gentle place to start.

Step Four: Attach a Safety Skill (Don't Skip This)

Wiring the MCP gives your agent the capability to trade. It does not give it the judgment to trade well. Before it signs a single order, attach a small persistent safety skill to your Instance from the Skills card. A sensible Kraken starting set:

• A per-trade notional cap and a rolling daily volume cap, so one bad decision can't scale into ten.
• A pair whitelist — only the markets you actually want traded (e.g. BTC/USD, ETH/USD), nothing the agent found trending.
• A rule to read account balance and open orders before every action, to avoid stacking or double-spending buying power.
• A ban on trading into thin or illiquid books, plus a minimum-liquidity floor so it won't enter a market it can't cleanly exit.
• A hard no on anything resembling withdrawals or transfers — belt and suspenders on top of the key scope from Step Two.

This lives in a skill, not a chat instruction, because chat instructions vanish when the agent's memory is compacted — persistent rules belong in persistent files. A skill survives every conversation, restart, and model swap. On a market that never closes, a rule left only in a chat window has all night to be forgotten. For the broader thinking behind these guardrails, our safety rails guide is the hub.

Step Five: Start Read-Only, Then Add Approvals

Give your agent a bounded first task that moves nothing. From the OpenClaw Direct dashboard, ask it to read your balances and open orders, pull current Kraken market context, and write a short brief on what it would trade if it were trading. This costs nothing, exercises the MCP end to end, and shows you how the agent reasons before it can act. Run that loop a few times. Sensible briefs are evidence the tools work; nonsense briefs catch a problem before any money moves.

When you graduate to live orders, keep a human in the loop. Wire a Telegram approval on every trade so you get a preview — pair, side, size, estimated price — with confirm and reject buttons before anything signs. Chatting with your agent turns out to be a natural place to catch a bad trade before it fills. Keep the first real order tiny — a small spot position in a blue-chip pair — then expand into scheduled strategies only once the plumbing has earned your trust.

What Happens When Something Goes Wrong

Every serious agentic setup has to answer one question: when your agent does something you don't want, how fast can you stop it? On Kraken you have two independent layers of protection, and they reinforce each other.

Your fastest stop is the OpenClaw Direct dashboard: suspend or terminate the Instance from any browser, on any device. Because the API key lives in the platform vault rather than on your laptop, killing the Instance means no process can sign a new order. Your second layer lives at Kraken itself — revoke or delete the API key from your account settings and its trading rights end immediately, no matter what any process tries. And because you scoped that key without withdrawal permission back in Step Two, even a fully compromised key can't move your balance off the exchange. Your funds stay custodied where they started.

The harder truth is what a kill switch can't undo. An order that already filled is filled; no dashboard button reverses a trade that cleared. This is why the per-trade and daily caps in your safety skill do the heavy lifting, and why the size of the balance the agent can reach — not the kill switch — is your real limit on the downside. The rule that matters most stays simple: don't put more into the trading account than you can lose without rearranging your life.

Where to Go From Here

The plumbing is finally first-party. Kraken shipped a CLI with a built-in MCP, so an OpenClaw agent can read one of crypto's largest exchanges, place trades through a key that can't drain you, and check in over Telegram before it acts — with whichever model you already trust. The remaining work — which pairs, how much size, how to grow trust over time — is the same work any trader does, and the right way to build those instincts is small stakes over a few weeks with read-only briefs and Telegram approvals. When you want to branch out, the tokenized equities hub and our Hyperliquid walkthrough are the natural next stops.

All of it depends on an agent that's actually online when the market is — which, in crypto, is always — and that keeps its API key somewhere safer than your laptop. That's why we built OpenClaw Direct: a dedicated Instance for every user, encrypted credential storage, 99.9% uptime, a full audit trail, a dashboard kill switch reachable from any browser, and Telegram approvals so a human signs off before anything fills. Free trial credits come with every subscription — enough to wire up Kraken end to end and watch your first few reads and orders fire.

---

Sources: Kraken Blog — Announcing the Kraken CLI, Kraken — API Documentation, and Kraken — xStocks (Tokenized Equities).