$ loading_
Set agent scopes and SOL caps, then propose, execute, and undo actions.
Copy the install command and let the AI configure it · recommended for beginners
Please install the "three.ws Autopilot" MCP server from askskill: Run: claude mcp add 'io-github-nirholas-autopilot-mcp' -- npx -y @three-ws/autopilot-mcp
Configure three.ws Autopilot with these rules: only allow on-chain actions within the specified project scope; set a daily SOL spend cap of 2 SOL; require proposal and approval before any high-risk action.
A clear autopilot policy configuration including scope, daily spend limit, and approval rules.
Based on the current task, first list the actions the agent plans to take, including estimated SOL cost and risk notes; do not execute anything until I confirm.
A proposed action list with cost and risk estimates for user approval.
Undo the most recent action the agent just executed, and report the undo result, affected resources, and any additional SOL cost.
A report describing the undo result, impact scope, and any related costs.
Track trending agents, coins, holder rankings, and site-wide activity updates.
Chat, generate text, create embeddings, and tokenize with IBM watsonx.ai Granite.
Use DashScope to access Qwen chat, embeddings, and model discovery.
Use IBM Granite pay-per-call for chat, code, embeddings, analysis, and forecasting.
Embed interactive live 3D avatars in agents or generate shareable embeds.
Equip AI agents with an x402 wallet to buy or sell services.
Check agent quotas, usage, invoices, receipts, and earnings in one place.
Browse boards, claim on-chain work, and post bounties through MCP.
Track an agent's portfolio value, PnL, balances, trades, and signed transfers.
Create 3D avatars, embeds, and manage agent memory and on-chain identity.
Generate 3D assets, rig models, and check agent reputation and market intel.
Analyze a tracked wallet’s P&L and trades for a specific token.