Unified API for running OpenCode, Claude Code, Codex agents
LiteLLM team solving agent SDK fragmentation with one query() function.
Developers building with multiple AI agent frameworks
LiteLLM · LangChain · LlamaIndex
lite-harness has one function - query()
import { query } from "@lite-harness/sdk";
for await (const message of query({ prompt: "Fix the failing test", options: { // swap harness between: "claude-agent", "openai-agents", "pi-ai" harness: "openai-agents", model: "gpt-5.5", }, })) { console.log(message); }
Local-first design — SQLite plus on-device ONNX embeddings — provides a small-footprint way to share operational lessons between agents, and automatic PII redaction is a thoughtful, tangible safeguard. The one-line publish/query API, cross-language SDKs (Python + TypeScript), and a planned FastAPI/Postgres+pgvector server show practical engineering trade-offs rather than vaporware.
AgentForge compresses common production patterns—token-aware rate limiting (token-bucket), retry+exponential backoff, prompt templates and cost tracking—into a tiny async core and lets you flip providers with one parameter. The multi-agent mesh and ReAct loop bits are the most interesting engineering bets here, and the repo includes benchmarks and a Streamlit demo, but it lives in a crowded space next to LangChain and similar toolkits so real differentiation will come from adoption and edge-case robustness.
Server-process architecture beats CLI wrappers for interactive agent sessions.
SQL queries over Claude/Copilot agent data; clever parsing, but tiny audience.
SDK-first messaging layer without forcing a framework or harness on your agents.
CogDB integration is the differentiator, but text-to-KG is a solved pattern.
