GitHub - LiteLLM-Labs/lite-harness: Unified API for running OpenCode, Claude Code, Codex agents

2 min read Original article ↗

Call all agent harnesses using the Claude Agent SDK format.

lite-harness manages:

  • One TypeScript and Python interface for multiple agent harnesses
  • Harness switching with harness, model switching with model
  • Claude Agent SDK-compatible streaming messages and errors

Preview: the SDK is not published to npm or PyPI yet. Clone this repo to try it. If you want a packaged release, please file an issue.

Discord

Setup (clone)

git clone https://github.com/LiteLLM-Labs/lite-harness.git
cd lite-harness

# install the backend server's deps once — the SDK auto-spawns it from the clone
npm install --prefix src/sdk/server

# pick a model — set the key for your provider:
export ANTHROPIC_API_KEY=sk-ant-...   # for harness "claude-code"
export OPENAI_API_KEY=sk-...          # for harness "codex"

TypeScript Usage

npm install --prefix src/sdk/typescript && npm run build --prefix src/sdk/typescript
import { query } from "@lite-harness/sdk";

const prompt = "Fix the failing test";

// Claude Code harness
for await (const message of query({
  prompt,
  options: { harness: "claude-code", model: "claude-opus-4-8" },
})) {
  console.log(message);
}

// Codex harness
for await (const message of query({
  prompt,
  options: { harness: "codex", model: "gpt-5.5" },
})) {
  console.log(message);
}

Python Usage

pip install -e src/sdk/python      # editable install of the client (Python 3.10+)
from lite_harness import query, AgentOptions

prompt = "Fix the failing test"

# Claude Code harness
async for message in query(
    prompt=prompt,
    options=AgentOptions(harness="claude-code", model="claude-opus-4-8"),
):
    print(message)

# Codex harness
async for message in query(
    prompt=prompt,
    options=AgentOptions(harness="codex", model="gpt-5.5"),
):
    print(message)

Supported Harnesses

See src/sdk/server/providers/ for the full list.

  • claude-code: Claude Agent SDK / Claude Code behavior. Upstream: Python, TypeScript.
  • codex: OpenAI Codex CLI behavior. Upstream: openai/codex.
  • pi-ai: Pi AI behavior.

With LiteLLM AI Gateway

Add LiteLLM AI Gateway when you want central keys, budgets, logs, fallbacks, and provider routing.

export LITELLM_API_BASE=https://litellm.your-company.com/v1
export LITELLM_API_KEY=sk-litellm-...
import { query } from "@lite-harness/sdk";

for await (const message of query({
  prompt: "Debug this production trace",
  options: {
    harness: "codex",
    model: "anthropic/claude-opus-4-8",
  },
})) {
  console.log(message);
}

Docs

SDK

License

MIT