Intent Cost Simulator

Demo

About

The problem it solves DeFi intent execution is expensive to reason about. Before submitting a sequence of transactions — swaps, loans, bridges, vault deposits — developers and solvers have no easy way to answer: *how much will this actually cost, and will it even succeed?* **Intent Cost Simulator** forks a live chain and replays your exact calldata against real EVM state, giving you: - **Per-step gas cost in USD** at current gas prices - **Cumulative flash liquidity required** — the capital a solver needs to front the sequence - **Revert detection** — whether each step fails and why, before you spend real gas - **Collateral scaffolding** — simulate against seeded balances to isolate logic errors from capital constraints Useful for protocol teams validating new intent flows, solvers pricing fill risk, and auditors stress-testing execution paths without touching mainnet. Use of AI tools and agents Claude and ChatGPT were used throughout for scaffolding — generating boilerplate for RPC client methods, the Anvil process management wrapper, and the reporter table layout. Useful for getting structure in place quickly, though most of the debugging and architecture decisions (fallback strategy, storage slot calculation, gas ceiling logic) required working through manually. Treated as a fast first draft, not a final answer.

How it's made

Challenges we ran into Two concrete ones: **1. `debug_traceCall` not universally available** Some chains (including ADI) don't support the `callTracer` tracer, causing the entire simulation to error mid-run rather than degrade gracefully. The fix was a fallback to `eth_call` when the tracer fails — preserving revert detection while flagging gas figures as estimated rather than measured. Still in progress for full parity. **2. Unicode panics in the terminal reporter** Column truncation was slicing strings at raw byte offsets, which panicked whenever a label contained multi-byte characters like `→` or `—`. The fix was replacing all direct byte slices with `char_indices()`-based truncation, and separating colorization from width calculation so ANSI escape bytes don't corrupt column math.