Runs the tools
It executes your compile and test commands and shows the command result on an expandable card.
Product
ERC 1155 smart contract work, one reviewed diff at a time
01
WaveClamp is an autonomous AI agent for Windows 10/11 that helps you write, test and review an ERC 1155 smart contract on a local chain or a test network. It reads your files, runs your own commands and puts every result in the chat as a card, so you see real output and not a description of it.
Shows every diff
Each edit appears as a diff that you accept or reject while Ask first is selected.
Plans before acting
Plan only writes the steps and runs nothing, so you can read the plan before anything changes.
What WaveClamp can do
- Filesread_file, write_file, append_file, edit_file, delete_file, make_dir, move_file, copy_file, list_dir, grep and glob work inside the chat's working folder, where your .sol files and tests live, and a counter shows files created, edited and read.
- Shellruns your own compile command and test command on Windows, in WSL or over SSH to a server you own, and judges success by the real exit code. Dangerous commands are blocked before they run.
- Webweb_search finds pages and http_get downloads one page, so the agent can read the public EIP text and the Solidity documentation while you work. Web search can optionally run over an anonymizing network.
- Screenshots and imagesocr reads text from a saved screenshot, such as an error dialog or a test report, and gen_image saves real image files, for instance placeholder artwork that your metadata JSON can point to.
- Orchestrationtask starts parallel sub-agents so separate contracts can be reviewed at the same time, run_on_model sends one hard step to a bigger model, and skill reuses a playbook such as a version-control skill for branches and commits.
- Approval modesAuto-run acts at once, Ask first pauses before actions that change something and shows the diff to accept, and Plan only writes a plan and runs nothing until you switch mode. Plans survive restarts.
- Cronruns a prompt on a schedule, for example a daily request to run your test command and list failing tests, so a regression shows up in the chat without a manual run.
- Modelsconnect a cloud AI provider through a standard API endpoint, run a model on your PC, or use the built-in model that is ready without your own endpoint. The Test button shows response time and the context window.
How a request becomes a reviewed change
You describe the task in the chat, the agent calls its tools, each call shows up as a card with the real result, and you accept or reject every diff that the agent proposes.
Find every function that mints tokens and list the ones without an access check.
- glob
**/*.sol3 matches - grep
_mint4 matches - read_file
GameItems.sol62 lines - edit_file
GameItems.sol+4 lines - Shell
your compile commandexit 0 - Shell
your test commandexit 0
waiting for your OK before the next stepRejectAccept
Four contract tasks, start to finish
Draft a game item contract
You want swords, potions and keys in one contract. The agent uses read_file to look at any existing .sol files, write_file to create the new contract with an SPDX comment and a pragma solidity line, and edit_file to add balances, a minting function and batch transfers. Making tokens moddable for a game means thinking about how items are described, and the metadata JSON behind uri(id) is the part a modder would edit. Each edit_file change arrives as a diff. Shell runs your compile command and your test command on a local chain, and you accept the diff only after you have read the exit code and the failing test names, if there are any.
Compare two standards
A design review starts with one question: one collection of unique tokens, or many token types in one contract? The agent uses web_search and http_get to read the public EIP text, then grep and read_file to compare it with the functions in your files. Put the ERC 721 Solidity file beside the multi-token one and ask for a table of differences, saved with write_file into a notes file. Questions about how to detect tokens that require an allowance reset belong to ERC-20, because ERC-1155 uses setApprovalForAll, which grants or revokes an operator for every token id at once. You review the notes before any code changes.
Review minting and access control
Whoever can call a minting function decides how many tokens exist, so access control is the first thing to review. The agent uses glob to find every .sol file and grep to list calls to _mint and _mintBatch with their line numbers. It then opens each function with read_file and reports the ones that any account can call. Ask first pauses before edit_file adds a modifier, and the diff shows the exact lines. The task tool starts sub-agents so separate contracts are reviewed in parallel. A review is not a guarantee. Run your test command, read the result, and get an independent security review before any real-value use.
Test on a local chain
Coins on a local chain or a test network have no value, so this is the place to practise, and anyone who wants to build their own ERC20 token faucet dapp can start there too. The agent uses Shell to run your command that starts a local test chain, then your compile command and your test command, and it reports each real exit code on a card. If a test fails, ocr can read a saved screenshot of the error and grep finds the event or function involved. You read the expanded card, accept or reject the proposed diff and run the tests again before you go on.
What WaveClamp does not do
WaveClamp is a desktop app, not a hosted service or a plugin. It is meant only for your own files, your PC and servers you are authorised to use, and you keep the work within them: file tools work in the chat's working folder, while Shell, web and SSH are not folder-bound. It gives no guarantee that code, scans or analysis are correct, and it gives no financial advice.
