Multi-agent systems that research, deploy, and operate DeFi tools on Fogo. Self-healing, memory-driven, no human in the loop. Real on-chain execution.
1 // Fogo agent: self-healing oracle + staking 2 const agent = await spawnAgent({ 3 name: 'fogofi-oracle', 4 chain: 'fogo', 5 memory: 'main.sqlite', 6 skills: [ 7 'pyth-sse', 'pool-prices', 8 'self-restart', 'tg-alert', 9 ], 10 cron: '*/5 * * * *', 11 }); 12 await agent.deploy(); 13 // running 24/7 on Fogo mainnet
Every agent has memory, skills, and a watchdog. They wake up on schedule, pick up where they left off, and only ask for help when stuck.
Each agent gets a name, a chain, persistent memory, and a list of skills. Skills are reusable building blocks: pyth-sse, swap, send, alert, deploy.
Cron runs the agent on its cadence. A separate watchdog checks that the process is alive every 2 min. Crash? Auto-restart. Token expired? Auto-refresh.
Persistent SQLite memory survives restarts. A reconciler runs every 15 min to fill any gaps the agent missed: failed orders, stuck states, low balances. Alerts to TG.
Active in production on Fogo mainnet. Each is a single PM2 process with its own memory, skills, and crash recovery.
Battle-tested primitives. Every new agent inherits these by default.
SQLite-backed memory survives restarts. Reload state where you left off.
Cron-based watchdog every 2 min. Dead process? Auto-restart with token refresh.
Groq for speed, Anthropic for reasoning, MiniMax fallback. Auto-select per task.
Direct Solana / Fogo / EVM execution. No middleware. spl-stake-pool, vortex, raw RPC.
Telegram, browser push, in-app notifications, SSE broadcast. Pick what fits.
Read-only API for safe metrics. CORS-open. Embed live data anywhere.
Need an autonomous bot for your project on Fogo or another SVM chain? Custom agents, integrations, and tooling. DM open.