workflowstacks
ai-agent

EasySteer

A Unified Framework for High-Performance and Extensible LLM Steering
293 stars24 forksJupyter NotebookUpdated 9/3/2026100% free · open source
What it does

EasySteer lets you add programmable “steering” layers to any LLM so you can bias its output (tone, style, safety, business rules) without retraining the model.

When to use it
  • •You need a fast, low‑latency way to enforce brand voice or compliance rules on LLM outputs.
  • •You want to experiment with multiple steering prompts (e.g., politeness, brevity) without fine‑tuning each model.
  • •Your product runs on a GPU and you need to swap steering modules at runtime for A/B testing.
Ready-to-paste prompt
curl -X POST http://localhost:8000/generate -H 'Content-Type: application/json' -d '{"messages":[{"role":"user","content":"What are the benefits of our premium plan?"}]}'
Heads up: EasySteer requires a CUDA‑enabled GPU and the same PyTorch version that the downloaded model was compiled for; forgetting to install `torch==2.2.0+cu121` (or matching your CUDA) will cause a runtime import error.
Saves to your device
Use with Claude
New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

✅ Light setup

Installs with a command or two; your AI agent can do it for you.

Try it instantly — no install
Claude Code
mkdir -p ~/.claude/skills/easysteer && curl -fsSL https://workflowstacks.com/api/skills/easysteer/claude-skill -o ~/.claude/skills/easysteer/SKILL.md
Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)
claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How EasySteer works
Codeflow
Free to inspect

EasySteer is a large Python project (~23k lines across 184 code files, plus 64 test files). Setup is light: runs with Docker. Last commit this month, Apache-2.0 license, has a test suite.

Size
Large codebase
~23k lines · 184 code files · ~3 h to skim
Setup
Light setup
Runs with Docker.
Runs on
Python · Docker
No API keys detected
Python 48%Jupyter Notebook 39%TypeScript 6%Vue 5%CSS 1%
Where to start reading
  1. 1
    README.md
    Start here — what it does and how to install it
  2. 2
    frontend/app.py
    Where the program starts running
  3. 3
    pyproject.toml
    Dependencies and the commands it exposes
  4. 4
    easysteer/__init__.py
    Inside easysteer/ — the main logic begins here
What's in each folder
easysteer/Core code — the actual logic103 files
frontend/Frontend / UI92 files
docs/Documentation17 files
replications/Folder57 files
experiment/Research & notebooks37 files
hf-space/Folder28 files
figures/Folder6 files
docker/Deployment / infrastructure4 files
READMEHas testsDocumentedCI checksDocker readyApache-2.0 licenseUpdated this month
Quick Actions
Details
Creator
ZJU-REAL
Language
Jupyter Notebook
Category
ai-agent
Published
7/6/2025

Are you the creator of this tool? Claim your listing → and earn 85% of every sale.

Related skills

More ai-agent tools founders pair with this one.

ai-agent★ 276,806
Track GitHub Stars with 996.ICU
Get a popular GitHub star counter for your startup, ideal.
ai-agent★ 264,191
Superpowers
An agentic skills framework & software development methodology that works.
ai-agent★ 244,424
Linux: Open Source Kernel
Get the Linux kernel source tree for custom development, ideal for startup founders in need of a flexible OS foundation
ai-agent★ 210,750
Vue
This is the repo for Vue 2. For Vue 3, go to https://github.com/vuejs/core
ai-agent★ 204,069
Improve Code with andrej-karpathy-skills
Get better code behavior with andrej-karpathy-skills, 165k+ GitHub stars, for founders using LLMs.
ai-agent★ 197,010
FreeDomain: Free Website Domain
Get a free domain with FreeDomain.