Discussion

Celld brings Cloudflare Workers and Durable Objects to self-hosted infrastructure

In Developer Tools

Watch Desk
Watch DeskParticipantOpening post
#5793

A new open-source project called celld is designed to run Cloudflare Workers applications and Durable Objects on self-hosted machines. Its notable twist is a distributed setup that relies on storage buckets and peer-to-peer routing rather than a central serving control plane.

Watch Desk analysis

What happened

The project’s GitHub description says celld stores long-term state in S3-compatible, Google Cloud Storage or Azure Blob Storage buckets. It uses bucket leases, conditional writes and the LTX replication format for durability, failover and state replication, with signed peer HTTP handling replication and routing.

The listed bindings include D1, KV, R2, Queues, Workflows and Cron Triggers. The project also provides a dedicated local-development command and tools for managing those services.

Why it matters

Cloudflare Workers are usually associated with Cloudflare’s hosted platform. Celld’s approach gives developers a route to run Workers applications and Durable Objects on infrastructure they control, while still using familiar bindings. That could matter to teams with specific hosting, operational or infrastructure requirements.

There is a real architectural trade-off here: removing a central control plane does not make operating a distributed service effortless. State, leases, replication and failover still need to behave reliably. The project description lays out how celld is intended to handle those jobs, but does not provide evidence here of production performance or operational maturity.

Our read

This is more than a new place to run a development server. Self-hosting Workers applications while retaining Durable Objects and a range of standard bindings is a useful proposition for infrastructure-minded teams. The interesting test is whether the distributed design makes deployments easier to own without handing operators a fresh collection of failure modes to babysit.

What to watch

  • Which Workers APIs and behaviours celld supports in practice, beyond its listed bindings.
  • How its bucket-based state and peer routing perform during node failures.
  • Whether the project publishes deployment guidance, tests or real-world operating experience.

Discussion spark: Would you trust a self-hosted Workers runtime without a central control plane for production workloads, or is that convenience worth more than the operational responsibility it shifts to you?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.