Hatchling

Hatchling is a command‑line chat client that brings the power of local large language models (LLMs) and the Model Context Protocol (MCP) into a single, interactive workflow. By integrating with Ollama and OpenAI backends, Hatchling lets developers run sophisticated LLMs on their own hardware while still enjoying the rich tool‑calling ecosystem defined by MCP. The result is a lightweight, extensible interface that turns any LLM into an intelligent agent capable of executing complex tool chains and citing external resources automatically.

The core problem Hatchling addresses is the gap between powerful local LLMs and the structured, reproducible tool‑calling capabilities of MCP. Many developers want to keep data on-premises for privacy or latency reasons, yet still need the ability to invoke external services—such as databases, APIs, or custom scripts—in a controlled manner. Hatchling solves this by acting as a unified front‑end: it forwards user prompts to the chosen LLM, interprets the model’s tool‑calling directives, and then dispatches those calls to registered MCP servers. The client also wraps longer chains of tool invocations, allowing an LLM to perform multi‑step reasoning without losing context or control.

Key features include:

• CLI chat interface that supports conversational memory, command shortcuts, and in‑chat tool toggling.
• LLM integration with both Ollama (local models) and OpenAI, giving developers the flexibility to choose performance versus cost.
• Tool execution wrapping that babysits LLMs, ensuring they can complete extended tool chains reliably.
• Automatic citation of the source software wrapped in MCP servers, so outputs carry provenance information without extra effort.
• Extensibility via Hatch packages, enabling teams to add custom MCP servers and tools directly from the command line.

In practice, Hatchling shines in scenarios where privacy, latency, or custom tooling are critical. For instance, a data scientist can use a local LLM to analyze proprietary datasets while invoking a bespoke statistical package through an MCP server, all within the same chat session. A DevOps engineer can chain together monitoring tools and deployment scripts, letting an LLM orchestrate complex rollouts. The planned integration with “Hatching! Biology” will further broaden use cases to bioinformatics, allowing researchers to run domain‑specific modeling software without leaving the chat.

By positioning itself as a thin, user‑friendly wrapper around MCP servers and LLMs, Hatchling streamlines AI workflows. Developers can focus on building domain tools while the client handles prompt routing, tool orchestration, and context management. This unified approach reduces friction, improves reproducibility, and empowers teams to harness the full potential of LLMs in their existing tool ecosystems.