DeFi Bots Series — Building an multi-strategy bot on Solana (Intro: Goals & Plan)

I’m reviving defi_server — an infra I maintained in my previous work with the Singapore startup behind taigo.ai. The old stack runs on Taigo MCP services and already knows how to talk to Solana. The goal now is to bolt on a new bot/ module and ship an autonomous scheduler that can put capital to work.

What I'm building

A multi-strategy bot that schedules and allocates across a few Solana venues using the existing defi_server plumbing:

• Meteora pools (DLMM) — Phase 1 focus
• Memecoins (Jupiter & pump.fun)
• xStocks (Solflare stocks)
• Perps (Vooi or a similar venue) — later, for risk hedging

Everything will initially live under bot/ and talk to the rest of the services (RPC, logging, state, alerting) that are already in the repo.

Phase 1: Meteora LP copy-trade

A simple first strategy: mirror a leader I follow on LP Agent.

• Leader portfolio: https://app.lpagent.io/portfolio?address=7KHx2Uc5qsqz652eXbu8Qtabi5KLxWJLgxFzcaBzP32i
• Bot behavior (v0):
  • Scan the leader’s recent transactions.
  • Detect a new DLMM position init.
  • Open the same pool on my side if it passes my guardrails (for now: assume yes).
• Important: I will not mirror their ongoing management. I’ll add my own rebalance, take-profit, and stop-loss logic. All trades/positions get recorded to Supabase, where I compute PnL.

Why copy first?

He’s had some profitable months. I’ll borrow his pool selection as a shortcut while I stand up the execution + risk framework. Once I’m tracking my own outcomes, I can layer on discovery signals and wean off copy inputs.

Series roadmap

1. (This post) Goals & plan; what “success” looks like.
2. Scanner: from raw tx history → “position intent” events (pool + position PDA).
3. Scheduler: cron-like loop, stateless vs. stateful triggers, backoff, idempotency.
4. LP executor: parse pool params, size the order, submit add-liquidity, record state.
5. Risk & lifecycle: TP/SL, rebalances, health checks, alerts.
6. PnL & reporting: Supabase schema, nightly marks, weekly digest.

Success criteria for Phase 1

• ✅ Bot detects the leader’s new DLMM position inits within my scan window.
• ✅ Bot opens the same pool with my own sizing.
• ✅ Every trade & state transition is stored in Supabase and can be replayed.
• ✅ I can publish transparent PnL as this series progresses.

Up next: the scanner — a lean pass over recent txs that identifies (pool, position PDA) without diving into complex bin-array reads. That’s Part 2.