Zero boilerplate. No reinventing LLM gateways, multi-tier memory, AST sandboxes, human approvals, or evaluation judges. Consumed directly from GitHub by AI coding agents.
Complete offline documentation, agent behavioral directives, and module references.
The behavioral rules, priority hierarchy ("Reuse first, build second"), and remote discovery contract for AI agents.
Detailed architecture breakdown of all 10 layers: Gateways, Memory, Tools, Orchestrator, Tracing, and CLI commands.
Exhaustive directory of all 40 modules with exported classes, functions, file paths, and exact Python usage examples.
Search by module name, ID, exported classes, or filter by architectural category.
How an AI agent (Cursor, Claude Code, GitHub Copilot, Codex, Antigravity) works in a different folder using this library.
In your new application folder, install the library directly from GitHub without cloning the repo:
# Using uv
uv add git+https://github.com/sh15aa/agentreusable.git
# Or pip
pip install git+https://github.com/sh15aa/agentreusable.gitTell any coding agent to reuse existing capabilities first instead of writing new boilerplate:
# In my-new-project/AGENTS.md
The project has access to: https://github.com/sh15aa/agentreusable
Module Registry: https://raw.githubusercontent.com/sh15aa/agentreusable/main/registry/modules.json
Rule: Reuse first, build second. Before scaffolding gateways, memory,
tools, or sandboxes, check registry/modules.json and import from `agentsreusable`.Give your AI agent this prompt in the new project:
"Build [your agent goal]. Check the reusable library at https://github.com/sh15aa/agentreusable and its registry at https://raw.githubusercontent.com/sh15aa/agentreusable/main/registry/modules.json. Identify existing modules and assemble them instead of writing boilerplate."
The agent imports standardized modules directly:
import asyncio
from agent_factory.factory import AgentFactory
from templates.research_agent import create_research_agent
async def main():
factory = AgentFactory()
agent = create_research_agent(factory)
result = await agent.run("Analyze latest AI agent architectures")
print(result)
if __name__ == "__main__":
asyncio.run(main())A battle-tested production blueprint separating concerns across the entire AI agent lifecycle.
Unique identity manifests, status lifecycle, environments, retry and cost policies.
Multi-provider routing by task complexity, streaming, and automatic fallback cascades.
Working scratchpad, multi-turn compaction, semantic vector search, episodic history & hybrid RAG.
Central registry, AST-checked Python sandbox, SSRF-guarded browser & document parser.
Dynamic planning engine, parallel DAG workflow executor, and supervisor multi-agent messaging.
Pause & resume approval queue intercepting HIGH & CRITICAL risk tool executions.
Pydantic structured output auto-repair, factual self-verification & citation provenance.
RBAC permission matrices, PII redaction (email/phone/SSN), and zero-leak secret isolation.
Golden test case datasets, LLM-as-Judge 1-5 rubric scoring & regression benchmark runners.
Hierarchical span tracing, security audit trail, operational metrics, and FastAPI SSE streaming.
Quick access to all connected platforms and tools