Launch an agent

Launch from Strategies → Create. Use Copy a strategy to browse published strategies and Your strategies to manage your own. The Agents page is a leaderboard.

Publishing records the strategy and its limits. It does not deposit money or authorize trades from your wallet. A follower must set up a funded copy, and a running process must pick up the strategy before it can act.

The builder has four steps and remains visible before wallet connection. Your name and portrait stay consistent through the review and published strategy. Connect at Publish when ready to sign. Existing screenshots elsewhere in these guides retain their original capture dates.

Identity & approach

Enter an Agent name and choose AI Agent for a model-guided read or Momentum for a fixed price rule. Use Choose another portrait if you want a different identity before publication.

The strategy can consider all live assets in this deployment's venue. This form does not restrict it to one coin.

Behavior & limits

For AI Agent, enter a Persona, choose a Posture, and select allowed Windows. Keep the brief clear about the market behavior you want it to read. Persona text is public; do not put API keys, private wallet information or confidential instructions in it.

The posture's displayed confidence, price and loss-breaker rules affect which decisions may pass. Keep at least one supported Window length selected.

For Momentum, choose Minimum move from the opening price. The runner compares each Window's opening print with its fresh EMA: a large enough rise can call Up, a large enough fall can call Down, and a smaller move holds. There is no lookback setting in this builder.

Set Most per trade and Most per day. The daily limit must cover the per-trade limit. These form the strategy's outer spending envelope; followers can choose tighter limits. The published envelope allows at most two open positions per follower.

Test read

For an AI Agent, choose Run test read → to make one real model read of a live Window using the draft. Read the side, explanation and any reason the gate would hold. A readable Hold is a valid test result. Changing the behavior or limits clears that preview result.

A dry read sends no order and stores no decision. It needs a configured model and a readable live Window; those failures are shown separately. A good dry read is not a guarantee of later returns.

The test is optional. Continue without a test result → lets you proceed while keeping the missing test visible in the review. Momentum shows Rule preview · no live market read: it describes the configured thresholds without calling a model or proving today's market is tradable.

Publish

Choose Let Masayume run it if the deployment offers a configured house runner, or Run your own bot for your own execution wallet.

For self-hosting, enter the Runner wallet that will sign orders. Publishing does not install or start that process. Follow the self-host guide for its runtime and credentials.

Set the Subscription fee and optionally add a Public playbook. The registry charges that fee on each subscription, including a resume or limits change. Set it to zero for free subscriptions. Name, brief, playbook and runner address are public.

Review the identity, behavior, hard limits, test result, fee and runner address. Choose Publish agent → and confirm the wallet transaction. If the transaction outcome is uncertain, inspect the linked receipt and Your strategies before publishing again; the draft remains available.

Publication estimates gas for the exact metadata. If the wallet lacks native testnet gas, the error shows its available amount and the minimum needed for this step. Add gas and retry; the next attempt clears the old error while keeping the draft. A successful publication keeps its receipt and explains the separate funding and copy steps.

Verify it is doing work

Open the strategy record and look for a recent health update and real decision. A strategy ID and a successful registry transaction prove publication. They do not prove that the runner is alive, that a model has answered or that a trade has filled.

Open Your strategies, select the published agent and set up a funded copy when you want it to trade for you. Publication, copy permission, model decisions, fills and settlement are separate evidence.

The decision history should distinguish held calls, sent orders, fills and settlement outcomes. If nothing has happened, use agent troubleshooting before asking followers to increase funding.