Skip to main content
Three tags, used consistently: verified (a real passing test or a real on-chain transaction you can check), built, untested end-to-end, or not started. Nothing below is rounded up.

Core build

On-chain policy

The practical consequence: the local ledger file is what actually gates real agent spending today. The on-chain cap is real and its logic is proven — but nothing currently routes production payments through it.

Dashboard

One caveat inside that last row, stated precisely: only the GET /xlm-price payment path has actually been click-tested by a human. The POST /summarize “Pay $0.02” button uses the same code path and would be expected to work, but “expected to” isn’t the same as verified — it hasn’t been.

Known loose ends, however small

  • TryItYourself.tsx sends a fixed placeholder string for the /summarize POST body rather than accepting typed input.
  • No settled transaction hash on record yet for a real /summarize payment via the dashboard.
  • Dashboard’s production JS bundle is ~1.5MB (Vite’s own build warning) — not code-split. Load-time concern, not a correctness one.
  • nymor-account’s constructor is intentionally rigid — one CallContract rule, one signer, one policy. A deliberate simplification for this project’s use case, not a general-purpose smart account.
  • CORS on nymor-resources is wide open (Access-Control-Allow-Origin: *) — correct for a public testnet demo, worth revisiting before any mainnet-facing claim.
  • No rate limiting on nymor-resources yet.
  • Building packages/policy requires a working host C/C++ toolchain and stellar-cli v25.2.0+ — real environment fragility was hit and worked around during development (see the repository’s commit history for specifics), not something a fresh cargo build alone resolves everywhere.