iAgent Autopilot — the first AI worker that actually trades on Injective, not just talks about it. Most AI tools read the chain and explain it. Autopilot acts: five specialized agents.
iAgent Autopilot is the execution and safety layer on top of the official Injective MCP Server — built for the Injective Solo AI Builder Sprint (May 2026).
Most AI agents on Injective read the chain or suggest trades. Autopilot executes within limits the operator defines: max notional, leverage, daily loss, allowed markets, and a one-click kill switch. Five specialized agents coordinate over a pub/sub pipeline (event → proposal → verdict → execution → audit), each persisted and streamed live to a Next.js workbench.
How AI is used: Groq Llama 3.3 70B (Watcher, Risk sanity check), Claude Sonnet 4.5 (Analyst, Auditor), Claude Haiku (strategy parse). Risk runs on a different model than the Analyst so the proposer never grades its own homework. The Executor has no LLM — it is the only component with MCP write access.
How Injective is integrated: All chain actions flow through MCP tools (trade_open, trade_close, transfer_send, balances, positions, etc.) on testnet. Public demo runs simulator + dry-run for judge safety; optional DEMO_REAL_TX produces an explorer-verifiable transfer_send for proof of execution.
<h3><strong>Shipped for the sprint</strong></h3><ul><li><p><span><strong>Five-agent runtime</strong></span> (FastAPI, SQLite, WebSocket) with deterministic risk gates, kill switch, and dry-run by default</p></li><li><p><span><strong>Operator dashboard</strong></span> on Vercel — strategy editor, agent fleet, decision pipeline, demo scenarios, live audit stream</p></li><li><p><span><strong>Public deploy</strong></span> — Render API + Vercel UI, API-key proxy, CORS, cold-start “waking engine” UX for judges</p></li><li><p><span><strong>Demo scenarios</strong></span> — Funding Reversion (full pipeline), Risk Block, Kill Switch; judge onboarding (pulse + hint)</p></li><li><p><span><strong>On-chain proof path</strong></span> — optional testnet <code data-inline="true" spellcheck="false">transfer_send</code> demo with explorer links</p></li><li><p><span><strong>Documentation</strong></span> — README (AI + Injective integration), <a href="http://DEPLOY.md">DEPLOY.md</a>, screenshots, <a rel="noopener noreferrer" target="_blank" href="https://youtu.be/4EFBRQVeDjw">90s demo</a></p></li><li><p><span><strong>Tests</strong></span> — 60+ runtime unit tests; dashboard production build verified</p></li></ul><h3><strong>Current limitations (honest)</strong></h3><ul><li><p>Render free tier <span><strong>cold-starts</strong></span> (~30s first load); dashboard retries automatically</p></li><li><p>Public deploy does <span><strong>not</strong></span> run Node MCP on Render — judges use <span><strong>simulator + dry-run</strong></span> (by design)</p></li><li><p>Live perp trading needs MCP + wallet on a VPS or local machine</p></li><li><p>Risk Block scenario depends on Analyst + Groq behavior; deterministic limits always apply</p></li></ul><h3><strong>Next steps (post-hackathon)</strong></h3><ol><li><p><span><strong>Mainnet hardening</strong></span> — stricter defaults, position caps, slippage checks</p></li><li><p><span><strong>Extend MCP tool registry</strong></span> — same pipeline for staking, bridge, limit orders (28 tools today)</p></li><li><p><span><strong>Strategy marketplace</strong></span> — import/export strategy JSON</p></li><li><p><span><strong>Backtesting</strong></span> — replay historical events through Watcher → Analyst → Risk without execution</p></li><li><p><span><strong>Multi-wallet</strong></span> — N strategies × N accounts with shared audit stream</p></li><li><p><span><strong>Hosted MCP</strong></span> — run Injective MCP beside the API for live demo without local setup</p><h3><strong>Value for the Injective ecosystem</strong></h3><p>Autopilot is <span><strong>infrastructure</strong></span>, not a one-off bot: a reusable governed execution shell for any agent that calls Injective MCP — so builders get safety, audit, and UX without rebuilding coordination from scratch.</p><hr><p><span><strong>It trades. You stay in control.</strong></span> — iAgent Autopilot: five AI agents, hard limits, Injective MCP, full audit trail.</p></li></ol>
<p>NOT YET BUT VERY SOON</p>