Lexichord

Lexichord is an AI orchestration tool designed specifically for technical writers. It coordinates multiple AI capabilities around a writer's actual workflow--terminology enforcement, audience analysis, style guide compliance, content structure--rather than wrapping everything in a single chatbot prompt and hoping for the best.

The Problem It Solves

Every major AI writing tool in 2026 follows the same basic pattern: give the user a text box, maybe let them paste in a style guide or select a "tone," and generate text. For casual writing, this is often good enough. For technical writing, it's not even close.

Technical writing has constraints that don't fit in a system prompt. Terminology databases with hundreds of approved terms, deprecated terms, and context-dependent usage rules ("use 'click' for mouse actions, 'tap' for touch actions, 'select' for generic instructions"). Audience-specific complexity rules that vary by document type. Document-type-specific formatting where procedures look different from conceptual overviews, which look different from troubleshooting guides. Content reuse requirements. Review cycle integration.

Pasting a 50-page style guide into a system prompt doesn't work. The model pays attention to some rules, ignores others, and has no way to handle conditional rules that depend on context it doesn't have.

How It Works (Conceptually)

Lexichord is an orchestration layer, not a model. It doesn't generate text directly; it coordinates other AI capabilities:

Drafting generates initial content using an LLM, but with structured context that includes document type, audience definition, relevant style guide rules, and approved terminology. The system prompt isn't a dumping ground; it's a carefully constructed context package.

Terminology checking compares generated content against a terminology database and flags non-compliant usage. Not keyword matching--semantic comparison that catches synonyms, near-misses, and the context-dependent terms that trip up every new writer on the team.

Audience analysis evaluates whether content matches the target audience's expected reading level, technical vocabulary, and assumed prior knowledge. A paragraph that's perfect for an API reference might be incomprehensible in a user guide. Style compliance checks against specific, enumerable rules: not just "does this sound like our brand?" but "does this follow rule 47b about capitalizing feature names in headings but not in body text?"

Content review produces structured feedback with specific, actionable suggestions. Not "this could be clearer" but "this sentence uses passive voice; consider 'Click Save' instead of 'The Save button should be clicked.'"

Each capability can use a different model, a different prompt structure, and different evaluation criteria. The orchestrator coordinates them and resolves conflicts. (What happens when the terminology checker and the style checker disagree? That's a more interesting problem than it sounds.)

Who It's For

Technical writers who work in structured environments with established style guides, terminology databases, and content types. Documentation teams that need consistency across multiple writers and AI-assisted workflows. Developer advocates producing technical content that must be accurate, audience-appropriate, and style-compliant.

It is not designed for casual writing, marketing copy, or creative content. Those are different problems with better-served tools.

Current Status

Lexichord is in active development. The product requirements specification and design specification were written first (naturally), and the orchestration architecture is being built in C#. The blog will cover the development process. The content plan includes posts on the orchestration manifesto, enterprise-grade AI tooling, the style guide problem, orchestration patterns in C#, and lessons learned.

Where to Find It

• GitHub: southpawriter02/lexichord (link will update when repo is public)
• Related glossary terms: Orchestration, System Prompt, AI Agent