Documentation intelligence for AI agents. Just as a museum docent guides visitors through exhibits, explaining context and connections, docent guides AI agents through your codebase.

Docent makes your agent better at understanding and improving documentation. No commands to learn—just configure once and talk to your agent naturally.

What Docent Does

Docent provides your agent with:

Project analysis - Understand languages, frameworks, structure
Quality assessment - Semantic evaluation, not pattern matching
Template library - ADRs, RFCs, specs, guides
Context gathering - Structured data for agent reasoning
Resource discovery - Runbooks, templates, standards via URI
Workflow prompts - Pre-defined procedures (RFC review, session recovery)

Quick Start

1. Configure Your Agent

Claude Code:

Add to ~/.claude.json:

{
  "mcpServers": {
    "docent": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@tnezdev/docent"],
      "env": {}
    }
  }
}

Then restart Claude Code.

Claude Desktop:

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "docent": {
      "command": "npx",
      "args": ["-y", "@tnezdev/docent"]
    }
  }
}

2. Talk to Your Agent

That's it. Just ask about documentation:

"How's my documentation?"
"Set up documentation structure"
"Create an ADR for switching to PostgreSQL"
"What should I document for this feature?"

Your agent will use docent automatically.

How It Works

When you ask your agent about documentation, it:

Uses docent to analyze your project
Gathers context about what exists
Reasons semantically (not pattern matching)
Provides contextual recommendations

You never invoke docent directly—it just makes your agent smarter.

MCP Capabilities

Docent exposes three types of MCP capabilities:

Tools

analyze - Project structure, languages, frameworks
audit - Semantic documentation assessment
list-templates - Available documentation templates
get-template - Fetch a specific template
capture-work - Append to work journal

Resources

Discoverable content via URI (docent://type/identifier):

docent://journal/current - Work journal for session continuity
docent://template/{type} - Documentation templates (adr, rfc, etc.)
docent://runbook/{name} - Operational procedures
docent://guide/{name} - Development guides
docent://standard/{type} - Project standards
docent://adr/{name} - Architecture decisions
docent://rfc/{name} - RFCs

Prompts

Pre-defined workflows agents can invoke:

resume-work - Session recovery with git, journal, TODOs
review-rfc - Multi-perspective RFC review
create-adr - Guided ADR creation
plan-feature - Feature planning workflow
research-topic - Structured research

Example Workflows

Session Recovery

You: "Help me resume work"

Agent: [invokes docent prompt 'resume-work']
Agent: [gathers journal, git status, commits, TODOs]
Agent: "You were implementing RFC-0005 (Enhanced MCP Architecture).
       Resources and prompts are complete. Next: update docs.
       3 uncommitted files. Ready to continue?"

RFC Review

You: "Review RFC-0005 from architecture perspective"

Agent: [invokes docent prompt 'review-rfc']
Agent: [reads RFC, analyzes project, applies criteria]
Agent: "APPROVE_WITH_CHANGES. Strong design but missing
       performance benchmarks. Recommend Phase 0 testing."

Assess Documentation Quality

You: "How's my documentation?"

Agent: [uses docent audit]
Agent: "Your documentation is solid (73/100).
       Strong architectural docs, missing testing guide.
       Want me to draft one based on your test structure?"

Follow Operational Runbooks

You: "Check CI/CD health"

Agent: [discovers docent://runbook/ci-cd-health-check]
Agent: [follows runbook procedures]
Agent: "CI/CD check complete. Test workflow passing.
       Lint workflow has 2 markdown errors. Fix them?"

Create New Documentation

You: "Document the decision to use PostgreSQL"

Agent: [uses docent get-template with type 'adr']
Agent: [creates docs/adr/adr-0004-postgresql.md]
Agent: "I've documented the PostgreSQL decision. Review?"

Target Audience

Solo developers using AI agents as part of their development workflow.

If you're using Claude, ChatGPT, or other AI coding assistants, docent helps them understand and improve your documentation.

Documentation

MCP Setup Guide - Detailed configuration
Architecture Overview - System design
RFCs - Proposed features

Development

# Clone and install (build happens automatically)
git clone https://github.com/tnez/docent.git
cd docent
npm install

# Add to ~/.claude.json for local development
{
  "mcpServers": {
    "docent": {
      "type": "stdio",
      "command": "/absolute/path/to/docent/bin/mcp-server.js",
      "args": [],
      "env": {}
    }
  }
}

# Restart Claude Code to load the MCP server