Six months of ethrex in production
Ethrex, our Ethereum execution client, has now spent six months running behind other people's mainnet validators. Staking operators put it in production, node-setup tools shipped it, and the programs that gate every client started gating this one. A production record is the one thing a client team can't grant itself, so this post keeps the receipts: every fact below links to someone else's repository, dashboard, or announcement.
Operators put stake on it
Serenita has listed ethrex on its operations dashboard since at least late January, with six months of continuous production as one of five execution clients behind its mainnet validators alongside Besu, Geth, Nethermind, and Reth. Serenita builds Vero, a validator client designed around client diversity, and backs that design with a five-client execution layer.
RockLogic, one of Lido's curated node operators, published a case study on its Lido infrastructure: "In production we run five different execution-layer clients (Geth, Nethermind, Besu, Reth, Ethrex) and six different consensus-layer clients (Nimbus, Prysm, Teku, Lighthouse, Lodestar, Grandine)." That puts ethrex validators inside Lido today, at the operator level.
On July 8, Stakely announced on X that it had incorporated ethrex into its Ethereum execution client setup, calling it "part of the operator work Ethereum needs more of: careful testing, monitoring, and real infrastructure support for client diversity." Stakely tests minority clients on Hoodi with more than 15,000 genesis validators before anything touches mainnet; ethrex has been through that pipeline and is now part of Stakely's mainnet's validator set.
The tooling ships it
The setup tools matter as much as the operators. Each of these projects writes and maintains its own integration.
eth-docker has shipped ethrex since October 2025 and has kept the integration current through 2026, tracking our releases in its changelog. Stereum merged its own ethrex service in January and made it selectable in the launcher UI with v2.4.6 in February 2026. Ethereum on ARM has packaged ethrex for ARM boards since October 2025, testnets included, and added a Grafana dashboard for it a month later. On the devnet side, EthPandaOps' Kurtosis ethereum-package has treated ethrex as a first-class client since July 2025, which is what lets us show up in every interop breakout.
In February 2026, clientdiversity.org added ethrex to its execution-client roster, marked stable.
Tested where every client is tested
In late April, we flew to Longyearbyen, Svalbard, for Soldøgn, our first interop as a participating client; the EF's recap covers the week. We were on every devnet that week, bal-devnet-4 through 6 and glamsterdam-devnet-0 through 2, implemented FOCIL during the week itself, and have kept pace since, through the v7.2.0 test fixtures in July.
Least Authority audited ethrex between January and May: 20 issues, 8 suggestions, and every issue was addressed. The client runs the EF execution-spec-tests and the cross-client Hive suites in CI, with daily reports.
The EF Applied Cryptography team's 2026 zkEVM program lists ethrex among its execution-layer teams, and its monthly updates track the work item by item, from execution-witness pre-serialization benchmarks in January to engine_newPayloadWithWitnessV5 in May. On Ethproofs, two of our clusters have been proving Ethereum mainnet blocks with ethrex's EVM as the guest program, on ZisK since December 2025, and on Airbender since early April.
The client kept moving underneath all of this: thirteen major releases since 2026 started, weekly since May. v9.0.0 opened the year with snap sync working across all networks and RocksDB as the default backend; v21.0.0 closed this stretch with Hegota fork support and initial EIP-8141 frame-transaction support. The 20x execution speedup and the snap-sync healing work from earlier in the year are what made "just run ethrex" a reasonable thing for an operator to say.
What's next
Adoption so far lives at the operator layer; the next layer up is the protocols. Getting ethrex into official client sets is the goal for the rest of the year. Our Lido presence today is one curated operator, and the path from there runs through more operators and the community staking side. StakeWise already accepts any client that speaks the Execution API, so its operators can run ethrex now.
The setup tools have room too. Sedge and DappNode have no ethrex option yet, and both follow the same shape as the tools that ship us, a service definition and a pinned image, and the Sedge PR is already open. The integrations that exist need tending as well. Stereum's service still pins an ethrex image from January, and keeping downstream packagers current falls on us.
On the diversity dashboards, we barely register, which is where every minority client starts. Moving that number is what all of the above is for.
The record so far
The record, as of July 2026: ethrex runs behind mainnet validators at three operators that chose it independently, one of them inside Lido's curated set. It passed its first external security audit with every issue addressed. It passes the same conformance and cross-client suites as every other client, in public, daily. clientdiversity.org marks it stable.
What the record doesn't show yet is scale. Ethrex's mainnet share is small, and its production history is measured in months rather than years. Reth, Besu, and Nethermind all carried the same caveats at this age; the way past them is the way they took: more operators and more time.
Our side of it is operational. The weekly release cadence is deliberate: every release is stable, and upgrading is a drop-in restart with no resync, so operators learn that keeping ethrex current is a non-event. When something does need fixing, it ships in the next release. Conformance and Hive results stay public, whether they flatter us or not. And the engineers who write the client answer directly on Telegram and GitHub.
If you run validators and want a second (or fifth) execution client, ethrex syncs mainnet in an hour and a half on 4-8 cores, 32 GB of RAM, and a 1 TB NVMe disk. Start with the getting started guide. Questions and feedback are welcome on GitHub, Telegram, or X.
