Skip to main content
A Solana transaction signature is opaque. Even fetching the raw transaction gives you a wall of instructions and program IDs that no LLM can reliably interpret. This endpoint returns a category (Swap, MemeCoin, Transfer), a plain-English summary, and the raw logs so your agent can decide what happened without hallucinating.
Paid endpoint. Pass your key in the x-api-key header. See Get an API key.
The current classifier uses log-substring heuristics (Jupiter program IDs, pump.fun markers, system-program transfer patterns). It’s fast and accurate for common cases. Complex composed transactions may fall back to Transfer.

Why agents use this

  • Logs alone confuse LLMs. Raw program invocations look like gibberish. A category label turns “what is this?” into an if-statement.
  • Audit trail for autonomous agents. Your bot signed a tx it built. Decode confirms it did what it intended.
  • Human-readable summaries for user-facing UI. No user wants to see Program log: Instruction: Route.
  • Post-mortem debugging. When a tx failed, the raw_logs + summary tell you why without opening Solscan.

Use cases

  • Auto-label wallet history for taxes. Loop through transactions output, decode each, group by category, export to your accountant.
  • Agent self-check. “I built a swap. Does the on-chain result confirm it as a Swap?” If category comes back Transfer, something went wrong.
  • Discord bot activity feed. User links their wallet, bot decodes new signatures every minute and posts “Bought BONK via Jupiter” instead of raw hashes.
  • Suspicious activity monitor. Flag MemeCoin category on a treasury wallet that should never touch memes.
  • Support ticket enrichment. User sends “my transaction did X, why?” Paste the signature, decode, respond with confidence.

Recipe: labeled activity feed for a Discord bot

activity-feed.ts
Combine with Token Profile: when decode-tx returns MemeCoin, immediately look up the token to see if it’s a rug your bot accidentally bought.