---
type: "post"
id: "the-harness-integrates-with-sglang-vllm-and-the-186be6"
slug: "the-harness-integrates-with-sglang-vllm-and-the-186be6"
title: "The harness integrates with SGLang , vLLM , and the experimental TokenSpeed engine through their registration hooks, with no runtime patching."
description: "EAGLE has described harness integration with SGLang, vLLM, and the experimental TokenSpeed engine."
retrieval_nugget: "EAGLE has described harness integration with SGLang, vLLM, and the experimental TokenSpeed engine."
published_at: "2026-10-21"
updated_at: "2026-10-21"
record_date: "2026-10-21"
date_kind: "published_at"
topics: ["harness-engineering","agent-runtime"]
entities: ["EAGLE","vLLM","SGLang","TokenSpeed","Qwen3.8-27B","EAGLE-4"]
related_patterns: ["harness-architecture-outlives-model-choice"]
novelty_label: "new_fact"
what_changed: "Adds that the EAGLE harness connects to SGLang, vLLM and experimental TokenSpeed through registration hooks, without runtime patching."
prior_refs: ["memory:16d2ef54-dbf5-86f2-a6aa-f4f4cef81db0","memory:529a238f-73fa-8d6a-a5fc-ec359bd058e0"]
source_url: "https://github.com/eagle-hq/vllm"
source_urls: ["https://github.com/eagle-hq/vllm"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:the-harness-integrates-with-sglang-vllm-and-the-186be6"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/the-harness-integrates-with-sglang-vllm-and-the-186be6/","markdown":"https://newruntime.com/posts/the-harness-integrates-with-sglang-vllm-and-the-186be6.md","json":"https://newruntime.com/posts/the-harness-integrates-with-sglang-vllm-and-the-186be6.json"}
---

# The harness integrates with SGLang , vLLM , and the experimental TokenSpeed engine through their registration hooks, with no runtime patching.

## Retrieval answer

EAGLE has described harness integration with SGLang, vLLM, and the experimental TokenSpeed engine.

EAGLE has described harness integration with SGLang, vLLM, and the experimental TokenSpeed engine. The integrations use each engine's registration hooks rather than runtime patching.

The reference pairing named with the integration is Qwen3.8-27B plus a published EAGLE-4 checkpoint. The public EAGLE vLLM repository identifies itself as an EAGLE-4 release fork of vLLM; its default release branch is `eagle/v0.1.0`, while `main` mirrors upstream. At the captured repository state, that release branch was 33 commits ahead of and 324 commits behind vLLM's upstream main branch.

The source does not provide benchmark results, throughput or latency figures, compatibility guarantees, installation instructions, or evidence that TokenSpeed is production-ready. It also does not establish which model configurations beyond the named reference pairing are supported.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Harness Engineering](https://newruntime.com/topics/harness-engineering.md): Explore the harness-engineering topic hub.
- topic hub: [Agent Runtime](https://newruntime.com/topics/agent-runtime.md): Explore the agent-runtime topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)