Find the right
payment agent.
A read-only MCP server for stellar-8004 agents that accept x402 micropayments. No key, no account.
59–91% of agent reviewers are Sybils.
59–91%
Sybil reviewers
3–15%
Live endpoints
0
Fields verified
7 checks passed · keyless · mainnet · healthy
One command. Any MCP client.
Node.js ≥ 22. No account, no API key, no wallet. Read-only by design.
-
One command registers the server through Claude Code’s own CLI, then performs a live MCP handshake.
npx -y stellar-agent-search@0.1.0 setup --client claude --scope user --handshake # The idempotent setup records this explicit stdio launch: # npx -y stellar-agent-search@0.1.0 mcp # Optional: install the skill your agent reads before calling anything npx skills add berkingurcan/stellar-agent-search --skill mcp
Re-run with --check --handshake to verify without changing config, or use --dry-run to preview the registration.
-
Same idempotent setup, targeting Codex’s TOML config — or register through Codex itself.
npx -y stellar-agent-search@0.1.0 setup --client codex --scope user --handshake # or let Codex write its own config: codex mcp add stellar-agent -- npx -y stellar-agent-search@0.1.0 mcp
Codex reads [mcp_servers.stellar-agent] from ~/.codex/config.toml — a JSON mcpServers block does nothing there. Asking setup for project scope prints the exact TOML without modifying anything.
-
One block in opencode.json — project root or ~/.config/opencode.
{ "mcp": { "stellar-agent": { "type": "local", "command": ["npx", "-y", "stellar-agent-search@0.1.0", "mcp"], "enabled": true } } }OpenCode uses an "mcp" root key with a command array, not the mcpServers shape. Restart the TUI after saving; all 13 tools appear read-only.
-
Project-scoped .cursor/mcp.json — conflicts are reported, never overwritten.
npx -y stellar-agent-search@0.1.0 setup --client cursor --scope project --handshake # The atomic config update records this explicit stdio launch: # npx -y stellar-agent-search@0.1.0 mcp
Existing matching config is left unchanged. Read-only limits server-side actions, but endpoint candidates remain self-declared, so keep client approvals aligned with your own policy.
-
Standard mcpServers JSON — drop the block into Hermes’ MCP config file.
{ "mcpServers": { "stellar-agent": { "command": "npx", "args": ["-y", "stellar-agent-search@0.1.0", "mcp"], "env": { "STELLAR_NETWORK": "mainnet" } } } }Same shape as Claude Code. The server stays read-only and keyless regardless of client.
-
Same mcpServers JSON in project .mcp.json — resources and slash prompts included.
{ "mcpServers": { "stellar-agent": { "command": "npx", "args": ["-y", "stellar-agent-search@0.1.0", "mcp"], "env": { "STELLAR_NETWORK": "mainnet" } } } }OpenClaw agents can call all 13 read-only tools, pin resources via @stellar-agent:stellar8004://…, and use slash prompts like /mcp__stellar-agent__find-and-vet-agent.
-
No agent required — the same binary is a plain terminal tool.
npx -y stellar-agent-search@0.1.0 find "web scraper" --x402 # discover npx -y stellar-agent-search@0.1.0 profile 10 # full profile for agent 10 npx -y stellar-agent-search@0.1.0 rank "scraping agents" --json npx -y stellar-agent-search@0.1.0 services --x402 # self-declared endpoint candidates npx -y stellar-agent-search@0.1.0 doctor # self-check
--json makes every command machine-readable.
Read-only. Keyless. By design.
- 01
Read-only and keyless
No signer, no write clients, no private keys anywhere under src/. STELLAR_PRIVATE_KEY is ignored on purpose and warned about on stderr. The only keyed actor in the repo is a standalone example, run under explicit human control.
- 02
stdout is JSON-RPC only
Every log and diagnostic goes to stderr, so the protocol stream is never corrupted by a stray print.
- 03
Agent text is data, never instructions
Names, descriptions, service labels and feedback tags live only in labeled selfDeclared slots of the structured output — sanitized (control, zero-width and bidi characters stripped) and length-bounded. Server-authored summary text interpolates only typed, enum or numeric values.
- 04
Degrade closed
Explorer reputation stays declared data. The bounded contract probe reports reachability, but sparse client indexes prevent a finite exhaustion proof; status is unavailable, verifiedFields is empty, and no summary-derived match is claimed.