diff --git a/op-wheel/README.md b/op-wheel/README.md
new file mode 100644
index 000000000000..ce5dfa298f14
--- /dev/null
+++ b/op-wheel/README.md
@@ -0,0 +1,92 @@
+# `op-wheel`
+
+Issues: [monorepo](https://github.com/ethereum-optimism/optimism/issues?q=is%3Aissue%20state%3Aopen%20label%3AA-op-wheel)
+
+Pull requests: [monorepo](https://github.com/ethereum-optimism/optimism/pulls?q=is%3Aopen+is%3Apr+label%3AA-op-wheel)
+
+`op-wheel` is a CLI tool to direct the engine one way or the other with DB cheats and Engine API routines.
+
+It was named the "wheel" because of two reasons:
+- Figuratively, it allows to steer the stack, an interface for a *driver* (like the op-node sub-component) to control the execution *engine* (e.g. op-geth).
+- Idiomatically, like the Unix wheel-bit and its slang origins: empower a user to execute restricted commands, or more generally just someone with great power or influence.
+
+## Quickstart
+
+### Cheat utils
+
+Cheating commands to modify a Geth database without corresponding in-protocol change.
+
+The `cheat` sub-command has sub-commands for interacting with the DB, making patches, and dumping debug data.
+
+Note that the validity of state-changes, as applied through patches,
+does not get checked until the block is re-processed.
+This can be used ot trick the node into things like hypothetical
+test-states or shadow-forks without diverging the block-hashes.
+
+To run:
+```bash
+go run ./op-wheel/cmd cheat --help
+```
+
+### Engine utils
+
+Engine API commands to build/reorg/rewind/finalize/copy blocks.
+
+Each sub-command dials the engine API endpoint (with provided JWT secret) and then runs the action.
+
+To run:
+```bash
+go run ./op-wheel/cmd engine --help
+```
+
+## Usage
+
+### Build from source
+
+```bash
+# from op-wheel dir:
+make op-wheel
+./bin/op-wheel --help
+```
+
+### Run from source
+
+```bash
+# from op-wheel dir:
+go run ./cmd --help
+```
+
+### Build docker image
+
+See `op-wheel` docker-bake target.
+
+## Product
+
+`op-wheel` is a tool for expert-users to perform advanced data recoveries, tests and overrides.
+This tool optimizes for reusability of these expert actions, to make them less error-prone.
+
+This is not part of a standard release / process, as this tool is not used commonly,
+and the end-user is expected to be familiar with building from source.
+
+Actions that are common enough to be used at least once by the average end-user should
+be part of the op-node or other standard op-stack release.
+
+## Design principles
+
+Design for an expert-user: this tool aims to provide full control over critical op-stack data
+such as the engine-API and database itself, without hiding important information.
+
+However, even as expert-user, wrong assumptions can be made.
+Defaults should aim to reduce errors, and leave the stack in a safe state to recover from.
+
+## Failure modes
+
+This tool is not used in the happy-path, but can be critical during expert-recovery of advanced failure modes.
+E.g. database recovery after Geth database corruption, or manual forkchoice overrides.
+Most importantly, each CLI command used for recovery aims to be verbose,
+and avoids leaving an inconsistent state after failed or interrupted recovery.
+
+## Testing
+
+This is a test-utility more than a production tool, and thus does currently not have test-coverage of its own.
+However, when it is used as tool during (dev/test) chain or node issues, usage does inform fixes/improvements.