What is the difference between Descrybe Legal Engine and the workflow pack?

Descrybe Legal Engine is the source-law research service and MCP tool layer. The workflow pack is a public set of assistant-facing research patterns that help an assistant use Descrybe in a structured way.

When should I use the Python SDK?

Use the Python SDK when you are building an app, script, local agent, firm tool, or coding-assistant project that needs authenticated access to Descrybe Legal Engine.

Do I need the workflow pack if I only use ChatGPT or Claude?

Not necessarily. The built-in setup guides are enough for many users. The workflow pack is useful when you want repeatable assistant behavior for research roadmaps, authority finding, or citation and quote auditing.

Is MCP the same as the Python SDK?

No. MCP is a protocol for connecting tools to compatible hosts. The Python SDK is a Python package and CLI for building apps, scripts, OAuth flows, token storage, and agent integrations that use Descrybe Legal Engine.

Can these tools provide legal advice?

No. These are legal research tools and workflow aids. Users still need to inspect sources and apply legal judgment before relying on the output.

Back to Resources

Workflow Pack, MCP Connector, or Python SDK: Which Descrybe Legal Engine Path Do I Need?

Descrybe Legal Engine is the source-law research service. The public workflow repository is the assistant-facing recipe book. The Python SDK is the app-building layer. This guide explains when to use the built-in ChatGPT, Claude, or Perplexity setup guides, when to use the workflow pack, and when to build with the SDK.

The short version

Descrybe now has several ways to use the same legal research layer. That is useful, but the names can blur together if you are deciding where to start.

The easiest distinction is this: Descrybe Legal Engine is the source-law research service; the workflow pack is the assistant-facing recipe book; the Python SDK is the app-building layer.

Use a built-in assistant connection when you just want to research

If your goal is to ask legal research questions in a familiar assistant, start with the setup guide for that assistant. ChatGPT, Claude, and Perplexity are the fastest paths when you want Descrybe tools available inside a conversation without building anything.

Use ChatGPT when you want Descrybe Legal Engine inside ChatGPT.
Use Claude when you want Descrybe Legal Engine through Claude connectors.
Use Perplexity when you want Descrybe as a custom MCP connector.
Use the Descrybe Platform when you want the full web workspace for research, review, saved work, and direct source inspection.

Use the workflow pack when you want repeatable assistant behavior

The public Descrybe Legal Research Workflows repository is for assistant-facing research patterns. It helps an assistant move from an issue, proposition, citation, or draft into a more structured legal research task.

This is useful when the assistant already has access to Descrybe Legal Engine, but you want a repeatable way to run a research roadmap, find authority, or audit citations and quoted language.

Research Roadmap for turning a plain-English issue into a source-grounded research path.
Authority Finder for organizing cases that support, limit, distinguish, or reject a proposition.
Citation and Quote Auditor for checking cited cases and quoted case language in a draft.
Claude-style and OpenAI-oriented packaging for different assistant surfaces.

Use the Python SDK when you are building the tool

The Descrybe Legal Engine Python SDK is for developers building apps, scripts, local agents, firm tools, or coding-assistant projects that need authenticated access to Descrybe Legal Engine.

The SDK includes the dle command-line tool, OAuth helpers, local token storage for single-user tools, refresh-token handling, a high-level LegalEngine wrapper, coding-agent instructions, and a browser-based example app.

Use MCP when the host already speaks MCP

MCP is the connection layer many assistant and agent surfaces use to call external tools. If the host already supports MCP, Descrybe Legal Engine can act as the legal research server behind that surface.

Most users do not need to think about MCP details when using ChatGPT, Claude, or Perplexity setup guides. Developers and workflow authors need to think about MCP when they are packaging tools, skills, or agent surfaces that call Descrybe Legal Engine directly.

A quick decision guide

I want to research in ChatGPT, Claude, or Perplexity: use the relevant setup guide.
I want better prompts and repeatable legal research workflows for an assistant: use the workflow repository.
I want to build a Python app, script, or local agent: use the Python SDK.
I want a full legal research workspace with saved work and direct source review: use Descrybe Platform.
I am building a team app: use the SDK with per-user OAuth, not one shared Descrybe login.

What all paths still have in common

All of these paths are meant to keep legal research close to sources. They can help search primary law, resolve citations, verify quoted language, check treatment, find citing authorities, and return source material for review.

They do not replace legal judgment. The final step is still to inspect important authorities, check jurisdiction and treatment, and decide how the sources apply.