Your program is failing on mainnet.
Do you know why?
WireTap is observability for deployed Solana programs. Decoded failures, drop diagnosis, release regressions, and compute-budget headroom — pushed to Telegram, pulled through MCP.
npx @wiretap-labs/cli init// try it — no signup
Paste any mainnet transaction signature. WireTap decodes why it failed against the program's IDL — the same answer you'd get in Telegram or through MCP.
Free public lookup, rate-limited. Register a program for alerts, drop classification, and history.
// three ways a Solana transaction dies
Solana failures come in three classes, and most tooling only sees the first.On-chain failures land and fail — but the explorer shows you an opaque hex code, not the IDL error your program actually threw. Dropped transactions never land at all — expired blockhash, exhausted compute, or an RPC that quietly lost them — and leave no record anywhere. Preflight failures die in simulation before they were ever sent. And the headline "~50% of Solana transactions fail" stat? Roughly 92% of those failures are bot spam. WireTap separates bot noise from real users so you see the failure rate that actually matters.
// nine questions, answered
This table is the product. Everything else is plumbing.
| # | Question | What you get | Where |
|---|---|---|---|
| Q1 | Why did this transaction fail? | Decoded, IDL-named error — SlippageExceeded, not 0x1771. | MCP why(sig) · TG /why <sig> |
| Q2 | Why didn't it land at all? | Drop classification: expired blockhash, CU exhaustion, RPC lag, or never-sent preflight. | MCP diagnose_drop(sig) |
| Q3 | Is it us, the RPC, or the network? | Your program's rate vs the ecosystem baseline vs a per-provider split. | MCP baseline(program) · TG /status |
| Q4 | Did our deploy break something? | Pre/post regression diff keyed to the deploy slot. | TG push · MCP release_diff(program) |
| Q5 | What are real users hitting most? | Top issues with bot traffic segmented out. | MCP top_issues(program) · TG digest |
| Q6 | Are we near the compute budget? | CU headroom per instruction, plus the trend line. | TG push · MCP cu_headroom(program) |
| Q7 | What happened to this user? | Wallet-scoped history across your program. | MCP wallet_history(program, wallet) |
| Q8 | Is our failure rate normal? | Landing Score percentile on the public leaderboard. | Leaderboard |
| Q9 | Which RPC provider actually lands? | Measured drop rate per provider, from your own traffic. | MCP provider_report(program) |
Telegram-first
Deploy regressions and CU alerts pushed where your team already is. Ask/why <sig> and get a decoded answer back.
MCP-native
Every question is an MCP tool. Point Claude or any MCP client at your programs and debug in plain language.
Open SDK
The client SDK is MIT-licensed — read exactly what runs in your app. The hosted product is flat per-program plans with a 14-day trial and a free public /why lookup.
npx @wiretap-labs/cli initZero to a decoded error in Telegram, on devnet, in under ten minutes.Follow the quickstart →