The modding scene around Hollow Knight: Silksong isn’t waiting for the starting gun. It hasn’t even been a “Skonging week,” as fans jokingly measure the passage of hype, and early experiments promising multiplayer are already surfacing across community channels. While Team Cherry has never announced online play for the long‑awaited sequel, modders are moving ahead with prototype builds and proof‑of‑concept tools aimed at getting multiple players on the same screen.
It’s a familiar story for this fandom: the original Hollow Knight inspired a robust ecosystem of tweaks, challenge modes, and unofficial co‑op experiments. But the speed at which Silksong efforts are appearing-some scarcely more than test beds-speaks to both the pent‑up demand and the technical confidence of a community that has spent years dissecting the series’ design. The projects remain early and unpolished, and their longevity will hinge on how the final game lands. For now, they’re a clear signal: when Silksong arrives, players won’t just be rushing in alone.
Modders move fast as community built cooperative prototypes surface well ahead of launch
In a burst of ingenuity that would make any speedrunner blush, community coders have spun up early cooperative play prototypes designed to stitch multiple Hornets into the same screen long before the official release lands. These builds-shared across mod hubs and invite-only test servers-lean on reverse-engineered hooks, animation sync layers, and lightweight scene managers to get two clients moving in lockstep. It’s rough, it’s provisional, and it’s already surprisingly ambitious, with contributors iterating nightly and pushing fixes for camera drift, enemy aggro, and UI collisions as they go.
- Shared camera and tethering systems to keep players framed during vertical climbs and boss arenas.
- Experimental netcode paths: delay-based for stability; early rollback trials for tighter inputs.
- Drop-in/out support, with emergency teleports and “ghost runback” options to reduce softlocks.
- Toggleable resource sync (currency, charms/tools, map state) for either shared or per-player progression.
- Local split-view prototypes vs. single-screen co-op, switching dynamically on player separation.
- Open documentation and test suites logging desyncs around iframes, ledge grabs, and parry windows.
As with any pre-release tinkering, expectations should be tempered: these are proof-of-concept mods with caveats around stability, platform support, and the shifting sands of pre-launch code. Maintainers stress compliance with developer guidelines and steer clear of assets or content distribution, focusing instead on modular loaders and reproducible tooling so the groundwork is ready when the real thing arrives. The upshot is clear: a volunteer QA pipeline is already forming-triage spreadsheets, nightly builds, issue trackers-and the conversation has moved from “Is co-op possible?” to “Which trade-offs get us the smoothest boss fights?” If momentum holds, day-one players could find a mature ecosystem of co-op options waiting, crafted by a community that’s building faster than most games can patch.
How the early multiplayer works tethering shared rooms limited enemy syncing and why desynchronization happens
Early co-op builds lean on a lightweight, peer-to-peer model with a host-authoritative core and a strict tether to keep players in the same screen or doorway cluster. The world is sliced into shared rooms; only the currently active room is networked, and a room transition typically requires both players near the threshold, preventing one client from streaming half the map alone. State replication focuses on the essentials: levers, doors, benches, geo orbs, and player transforms. Enemies are broadcast with limited syncing-usually health, phase flags, and a small state ID-while pathfinding, animation blends, and micro-physics run locally. That keeps packets small and latency tolerable, but it also means what you see is a negotiated approximation rather than a perfect mirror.
- Tether behavior: a soft leash escalates to a snap-back or warp if a player strays, preserving camera and culling sanity.
- Room lockstep: transitions are gated; interactables fire once via host RPCs to avoid double triggers.
- Enemy updates: periodic health/state deltas, not per-frame AI; projectiles and hitboxes are often client-predicted.
- Bandwidth discipline: compress transforms, throttle sends, and coalesce events to survive shaky Wi‑Fi.
Desynchronization creeps in wherever determinism ends. Different frame rates, floating‑point drift, and non-aligned RNG seeds push clients onto divergent AI branches; add latency, packet loss, and tick-rate mismatches, and you get enemies taking a hit on one screen but not the other. Local-only collision resolution makes phantom hits and rubber-banding likely, and long boss phases magnify tiny timing errors until health totals disagree. Mod hooks can’t yet intercept every damage pathway or animation notify, so even clean snapshots may reintroduce drift as soon as players resume combat.
- Symptoms: teleporting partners, duplicated pickups, delayed stag transitions, desynced stagger windows.
- Mitigations: periodic host snapshots, room-change hard resyncs, host-only damage authority, input delay or lag compensation, and CRC checks on entity tables.
- Trade-off: tighter tether and fewer synced entities equal stability-but reduce the illusion of seamless co-op.
Risks to weigh unsupported hooks platform policies save corruption and inevitable breakage after patches
Early co-op hooks are exciting, but they’re still unsanctioned scaffolding built on a shifting codebase. On PC storefronts, and especially on consoles, that matters: injecting DLLs, memory patching, or side-loading frameworks can collide with platform policies, trigger automated moderation, or simply get you locked out of online features. Even without overt anti-cheat, behavioral heuristics exist. The legal gray is as real as the technical one-mod loaders that touch networking stack behavior or entangle with cloud systems can look indistinguishable from exploits.
- Back up your save folder and disable cloud sync before testing.
- Keep experimentation to PC; console tinkering risks account action under terms of service.
- Use private sessions only; avoid matchmaking that could trip automated checks.
- Test on a secondary profile in a sandboxed install, not your main account.
- Read the mod’s license and telemetry notes; avoid builds that beacon to unknown endpoints.
Technically, the sharpest edge is brittleness. Patches shift function addresses, alter serialization, and invalidate assumptions that makeshift multiplayer relies on. The result can be save corruption, chronic desyncs, soft-locks, and “ghost” progress after a hotfix lands. Auto-updaters make this worse by moving the target mid-session. Treat each update as a new game until maintainers catch up-and assume your progress is at risk until proven otherwise.
- Wait for maintainers to publish compatibility updates after each patch.
- Pin to a known-good build ID and pause auto-updates before playing modded.
- Verify your loadout with a checksum/mod manager before opening a save.
- Keep separate save branches for modded and vanilla runs.
- At the first sign of trouble-desyncs, missing UI, log spam-exit, restore backups, and wait.
Practical advice for safe testing use trusted repositories isolate installs back up saves play offline and be patient
Early builds of multiplayer mods are exciting, but the safest testing starts with provenance and transparency. Treat every download like a newsroom source: verify, corroborate, and keep receipts. Favor communities with visible maintainers, public changelogs, and reproducible releases. Watch for mismatched file sizes, vague permissions, and packs that bundle unrelated binaries-classic tells of a bad scoop.
- Use trusted repositories: Prefer well-known mod hubs or the developer’s official mirrors; check release notes, hashes, and maintainer history.
- Cross-check feedback: Scan comments and issue trackers for crash reports, malware flags, and version conflicts before installing.
- Inspect before you run: Unpack to a sandbox, scan archives, and avoid installers that demand elevated privileges without a clear reason.
Contain the blast radius. Build a dedicated test environment so experiments don’t spill into your main save or live setup, and approach each session like a controlled trial. Keep your connection on a short leash, and give mod authors time to iterate-early netcode will wobble before it walks.
- Isolate installs: Duplicate the game directory or use a separate library folder; disable cloud sync for the test build to prevent cross-contamination.
- Back up saves: Copy your save folder to an external location and archive it; restore points save hours when a patch goes sideways.
- Play offline first: Switch your client to offline mode or firewall the executable to avoid version drift and accidental updates mid-session.
- Be patient and methodical: Expect desyncs, rollbacks, and hotfix churn; pin stable mod versions, update one component at a time, and report reproducible bugs with logs.
However this shakes out, the takeaway is clear: the appetite for Silksong is so intense that the community isn’t waiting for official word before trying to share Hallownest with a friend. Early multiplayer mods are always more proof-of-concept than polished feature-expect desyncs, crashes, and fast-moving updates-and users should exercise the usual caution: verify downloads, back up saves, and understand the risks that come with unofficial tools.
What matters, for now, is the signal. In less than a “Skonging week,” modders have mapped enough of the game’s systems to sketch out co-op, a reminder of just how quickly hobbyists can extend a precision platformer in unexpected directions. Whether Team Cherry ultimately embraces or discourages this experimentation will shape what’s possible next. We’ll be watching how these projects evolve, which ones stabilize, and whether any official guidance on mod support appears. Until then, the message from the scene is unmistakable: Silksong isn’t out, but its community is already playing together.
