AI agents are getting smarter, but they still can't tell who to trust. Feed an agent a Nostr pubkey and it has no idea if that account is a real developer with 5 years of contributions or a bot created 10 minutes ago.
The maximumsats-mcp server gives your agent 10 Web of Trust tools that answer these questions instantly. It indexes 51,000+ Nostr pubkeys using PageRank over the follow graph.
Setup: 30 Seconds
npx -y maximumsats-mcpThat's it. Your MCP client (Claude, Cline, Cursor, etc.) now has 10 WoT tools. No API keys needed — 50 free requests per day, then L402 micropayments kick in at 3-10 sats per query.
What Your Agent Can Do
Check a trust score:
wot_score({ pubkey: "npub1..." })
→ { score: 18, rank: 3, percentile: 99.99, followers: 1358 }Detect sybil accounts:
wot_sybil_check({ pubkey: "npub1..." })
→ { sybil_score: 64, classification: "likely_genuine", signals: [...] }Find trust paths between accounts:
wot_trust_path({ from: "npub1...", to: "npub1..." })
→ { connected: true, paths: [...], combined_trust: 0.216 }Analyze follow list quality:
wot_follow_quality({ pubkey: "npub1..." })
→ { quality_score: 39, ghost_ratio: 0.38, suggestions: [...] }Why This Matters for Agents
If your agent interacts with Nostr (posting, reading, transacting), it needs to know who to trust. A few use cases:
- Spam filtering: Before showing a DM or mention, check the sender's WoT score. Below 5? Probably spam.
- Payment decisions: Before sending sats to someone, verify their trust score and sybil classification.
- Content curation: Rank posts by author trust score instead of chronological order.
- Social verification: Check if two accounts have trust paths connecting them — no path means no social proof.
The Full Tool List
| Tool | What it does |
|---|---|
wot_score | PageRank trust score (0-100) |
wot_sybil_check | 5-signal sybil detection |
wot_trust_circle | Mutual follow analysis |
wot_follow_quality | Follow list health check |
wot_trust_path | Find trust paths between pubkeys |
wot_network_health | Graph-wide stats (51K nodes) |
wot_anomaly_detection | Flag suspicious patterns |
wot_link_prediction | "People you may know" |
wot_compare_providers | Cross-provider score verification |
wot_influence_analysis | Simulate follow/unfollow impact |
L402 Payments: Your Agent Pays Per Query
After 50 free requests, the API returns HTTP 402 with a Lightning invoice. Your agent pays 3-10 sats via the Alby MCP server (which has a fetch_l402 tool), and retries with the payment hash. The whole flow is automated — no human intervention.
Agent calls wot_score → 402 + invoice
Agent calls alby.pay_invoice → paid
Agent retries wot_score with payment_hash → dataThis is the L402 protocol in action: machine-to-machine micropayments over Lightning. Your agent becomes an autonomous economic actor.
Try It Live
The interactive demo at wot.klabo.world/demo shows all 13 analysis cards. Paste any Nostr pubkey and explore the data your agent would get.
API docs: wot.klabo.world/docs | OpenAPI spec: wot.klabo.world/openapi.json | Swagger UI: wot.klabo.world/swagger