ToolProduction ReadyMCP

zypher-agent

Name: zypher-agent
Rating: 3.0 (335 reviews)
Author: corespeed-io

by corespeed-io

Minimal TypeScript framework for configurable AI agents and multi-agent flows

TypeScript

Updated Jul 12, 2026

335

Stars

Forks

View on GitHub

How It Works

Provides a minimal TypeScript framework for building AI agents with explicit control over tools, providers, and execution flow. Structures agents as composable skills and runtime primitives so you can swap LLM providers, tool integrations, and execution strategies without changing agent logic. Includes an MCP client and supports swapping tool integrations via the Agent Registry Pattern for flexible orchestration and integration.

Need implementation help?

Expert guidance available

Key Benefits

As agents become more autonomous, reproducible control over their toolset and providers is essential for evaluating reliability and attribution. Zypher-agent makes agent behavior explicit and configurable, which helps teams record agent decisions, reproduce runs, and compare outcomes across providers — foundational needs for agent-to-agent evaluation and building an agent track record. Until now many agent frameworks hide provider and tool plumbing, making consistent A2A evaluation and trust auditing harder.

Best For

Developers building production or pre-production agents who need explicit control over LLM providers, tools, and multi-agent message flows. This is supported by the Orchestrator-Worker Pattern to facilitate coordinated interactions across agents.

Use Cases

Prototype agents that need provider-agnostic behavior for A/Bing LLMs
Record and reproduce agent runs to establish an agent track record for evaluation
Connect agents across processes using MCP for multi-agent orchestration and testing