HelpNestBeta
AI-first

Your help center, built for AI agents

HelpNest exposes your knowledge base through llms.txt, MCP, and A2A — so any AI agent can discover, search, and query your help articles without custom integration.

MCP on npm →Source on GitHub

Three protocols

Every way an AI agent can talk to your help center

From simple file fetches to stateful multi-turn conversations — choose the integration level that fits your use case.

llms.txt

Discovery

Here's what we know

A static text file at /llms.txt that describes your help center in a format any LLM can read. Zero integration required — just fetch and parse.

MCP

Tools

Call these tools

Model Context Protocol gives AI assistants like Claude structured tool access to search articles, fetch content, and ask questions — all stateless.

A2A

Collaboration

Let's collaborate

Agent-to-Agent protocol enables stateful, multi-turn conversations between AI agents and your help center — with task tracking and streaming.

terminal

# Discover what your help center knows

$curl https://demo.helpnest.cloud/helpnest/help/llms.txt

 

# HelpNest Help Center

 

> Support docs, guides, and answers for HelpNest.

 

This help center contains 17 articles across 6 collections.

 

## Collections

- [Getting Started](.../getting-started)

- What is HelpNest?

- Quick Start (Local Development)

- Environment Variables

- [Self-Hosting](.../self-hosting)

- [Using HelpNest](.../using-helpnest)

- [Integrations](.../integrations)

- [About](.../about)

- [Migration](.../migration)

 

## Full Content

For complete knowledge base content, see:

[llms-full.txt](/helpnest/help/llms-full.txt)

llms.txt

The simplest way for LLMs to understand your help center

A plain-text file at /llms.txt that any model can fetch and parse. No authentication, no SDK, no integration work.

/llms.txt

A concise overview of your help center — collections, article counts, available API endpoints, and protocol support. Designed for quick context injection.

  • Auto-generated from your live content
  • Collections with article counts
  • Links to search, MCP, and A2A endpoints
  • Updates automatically as you publish
/llms-full.txt

The full content of every published article, concatenated into a single file. Ideal for RAG pipelines or bulk ingestion into a vector store.

  • Every published article in full Markdown
  • Structured with clear article boundaries
  • Perfect for embedding pipelines
  • Single fetch — no pagination needed

Model Context Protocol

Give Claude and other assistants direct access to your help center

Add HelpNest as an MCP server and AI assistants can search, read, and query your knowledge base using structured tools.

# Add HelpNest MCP to Claude Code

$claude mcp add helpnest \

-e HELPNEST_API_KEY=hn_xxx \

-e HELPNEST_WORKSPACE=your-workspace \

-e HELPNEST_BASE_URL=https://dashboard.helpnest.cloud \

-- npx -y @helpnest/mcp

Available tools

search_articles

Full-text search across your knowledge base

get_article

Fetch a single article by ID or slug

list_collections

Browse the collection hierarchy

ask_question

AI-powered Q&A over your content

@helpnest/mcp on npm →

Agent skills

Teach any coding agent to write your help center

Skills package the workflow, tone, and safety rules for authoring HelpNest content. Install once with the skills CLI and your agent knows how to draft, refresh, and publish articles through the MCP server — no prompt engineering required.

skills.sh

# Install into every agent you have

$npx skills add babble-open-source/helpnest-skills

 

✓ Detected Claude Code, Codex, Cursor

✓ Wrote helpnest-author to 3 agents

 

# Or install globally / target one agent

$npx skills add babble-open-source/helpnest-skills -g

Available skills

helpnest-authorskill

Author and refresh help articles

Packages the full authoring workflow — workspace verification, collection lookup, idempotent create-or-update, and draft-vs-publish safety — so your agent writes customer-ready articles straight through the MCP server.

  • Grounds every article in your real UI labels
  • Idempotent create-or-update — safe to re-run
  • Defaults to drafts for review before you publish
babble-open-source/helpnest-skills →

Agent-to-Agent

Stateful conversations between agents and your help center

The A2A protocol lets external agents discover capabilities, send messages, and track tasks — with optional SSE streaming for real-time responses.

agent card discovery

# Discover the agent's capabilities

$curl https://help.acme.com/.well-known/agent-card.json

 

{

"name": "Acme Help Center",

"url": "https://help.acme.com",

"capabilities": {

"streaming": true,

},

"skills": [

"search", "ask", "get_article"

]

}

message/send

# Send a question via JSON-RPC

$curl -X POST /api/a2a \

-H "Content-Type: application/json" \

-d '{

"jsonrpc": "2.0",

"method": "message/send",

"params": { "message": {

"role": "user",

"parts": [{ "text": "How do I reset my password?" }]

}}

}'}

message/send

Send a message and get a complete response

message/send-streaming

SSE stream for real-time responses

tasks/get & tasks/cancel

Track and manage long-running tasks

Agent-optimized API

REST endpoints designed for AI consumption

Standard REST API with Markdown output, bulk export, and incremental sync — everything an agent needs to stay current.

GET

/api/articles?format=markdown

Returns articles in clean Markdown — ideal for LLM context

GET

/api/articles/export

Bulk export all articles as a structured JSON payload

GET

/api/articles/changes?since=ISO

Incremental sync — only articles modified since a timestamp

Make your help center agent-ready

Every HelpNest instance ships with llms.txt, MCP, and A2A support out of the box. No configuration needed — just deploy and your knowledge base is accessible to any AI agent.

MCP on npm →Source code