WrenOS
WrenOS OpenClaw
Easy. Safe. Deploy in one-click.
Your agent's trading stack for Solana memecoins.
Open source trading agents with private inference and safe defaults.
SELF-HOSTED
Advanced. Full control.
FROM SIGNAL TO TRADE
Agents that trade on evidence, not vibes.
This isn't a black-box that trades for you. It's a structured pipeline you control with research, scoring, backtesting, and execution as separate stages you can inspect, tune, and override. The system gives you tooling and guardrails. The strategy is yours.
A research agent scans on-chain wallet flows, social signals, and market feeds to surface a ranked watchlist of tokens worth looking at. A trading agent then evaluates each candidate against your strategy rules and regression gate.
Discover
→
Score
→
Validate
→
Execute
The goal is simple: give you the tooling, guardrails, and workflow to build conviction from evidence. Execute only what survives your process.
Ready to start?
WrenOS + RAILWAY
Recommended
Deploy in one click
Deploy on Railway with pre-configured profiles. Connect to Telegram, set your env vars and you're trading in minutes.
SELF-HOSTED
Advanced
Already running OpenClaw?
Add WrenOS to your existing setup. Clone the repo, install profiles, wire MCP servers. Full control over every config.
BUILT ON OPENCLAW
Real infrastructure, not a whitepaper.
7
Agent profiles
310+
MCP tools
17
Vendored skills
WrenOS runs as an inspectable OpenClaw system. Agent coordination, handoffs, and execution controls are explicit and auditable. Research and strategy evaluation are separated by design. Execution defaults to paper mode.
Every token earns its way in: discovered, scored, stress-tested, or rejected.
Strategy inference routes through Speakeasy for private models with zero content logging. Your agent can reason about MEV strategies, analyze arbitrage patterns, and evaluate opportunities that frontier models refuse to discuss. Four model routes: deepseek-v3.2 for research, qwen3-thinking for deep reasoning, qwen3-coder for refinement, venice-uncensored for unrestricted analysis.
Fail-safe controls are built in: confidence tiers automatically degrade as data quality drops, reducing risk before execution. Everything is open source under Apache-2.0, so you can inspect the full stack end to end.
TOOLKIT
310+ tools across 4 servers, 17 skills, and 2 agent packs.
Everything vendored. Nothing to configure. Works out of the box.
49 tools
MARKET INTELLIGENCE
Real-time crypto news, DeFi analytics, whale alerts, funding rates, and market sentiment — all through a single MCP server. Your agent gets market context without scraping or API key juggling.
~200 tools
ON-CHAIN DATA
Full-stack crypto data layer covering EVM chains, DeFi protocols, DEX analytics, and wallet tracking. The broadest single MCP server in the stack.
36 tools
BONDING CURVES
Direct interface to pump.fun — quote trades, build token creation transactions, track graduation progress, manage fees. Purpose-built for meme token lifecycle.
~25 tools
SOLANA RPC
Wallet balances, transaction history, token metadata, and asset search via Helius. No middleman between your agent and the chain.
17 vendored
SKILLS
Curated knowledge covering DeFi, security, Jupiter integration, whale watching, MEV protection, and more. Each skill loads when relevant — no training required.
2 ready-made
AGENT PACKS
Dual Agent Pack for research-to-trading handoff. Meme Discovery Pack for the full pipeline — composite scoring, zoo-style strategy management, confidence tiers built in.
Built for agents and humans.
Point your agent at the skill file and it self-configures. Or read the quickstart and deploy on Railway in two minutes. Everything is discoverable, inspectable, and documented.
SKILL.md
Root skill file — point your agent here to self-configure
QUICK START
One-command install or Railway deploy — two minutes
GITHUB
Browse source, profiles, packs, skills, and adapters
Discoverable via <link rel="skills" href="/SKILL.md" /> in the HTML head.