A wallet your agent
can actually spend from
USDC. Hard spend caps. Allowlisted scope. A signed receipt for every payment. Drop into Claude, Cursor, LangChain, CrewAI, AutoGen, or LangGraph — same install, same wallet.
Hard spend limits
Per-call, per-day, per-month. All enforced client-side before signing — payments that would breach the cap never reach the chain. Set in env, config, or the dashboard.
Allowlisted hosts
Exact hosts, wildcards, or path patterns: api.openai.com, *.cloudflare.com, docs.example.com/research/*. Anything outside the allowlist is rejected before signing.
Signed audit trail
Every payment generates a receipt with tx hash, URL, amount, timestamp. Verifiable on Basescan. Filterable by host, date, agent.
Your key, your machine
Wallet generated locally on first run. Key never leaves the agent's machine. Xenarch never sees, signs, or holds anything. No ETH required.
MCP clients
claude mcp add xenarch
mcpServers config
npx @xenarch/agent-mcp
SDKs & adapters
pip install xenarch
xenarch[langchain]
xenarch[crewai]
xenarch[autogen], xenarch[langgraph]
npx @xenarch/agent-mcp. For Python agents: pip install xenarch. For frameworks: xenarch[langchain], xenarch[crewai], xenarch[autogen], xenarch[langgraph]. First run generates a wallet locally.~/.xenarch/wallet.json by default. Override with the XENARCH_PRIVATE_KEY env var. Xenarch never sees or stores it.