Helpjuice MCP Server
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Helpjuice MCP Server search the knowledge base for how SSO is set up"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Helpjuice MCP Server
A Model Context Protocol server for Helpjuice. It lets Claude, Cursor, ChatGPT and other AI agents search and read your knowledge base, draft and update articles, and audit stale content.
Unofficial. This is a community project and is not affiliated with Helpjuice. It was built from Helpjuice's public API v3 docs.
What you can ask your agent
"Search our knowledge base for how SSO is set up and answer this customer's question, with a link to the article."
"Which published articles in Billing haven't been touched in a year? Give me a table, oldest first."
"Turn this Slack thread into a draft help article in the Integrations category. Don't publish it."
"Our pricing page changed. Find every article that mentions the old 'Starter' plan and propose edits."
"What did the team publish or edit this week?"
Related MCP server: kb-intelligence
Tools
Tool | What it does | Writes? |
| Full-text search with clean text samples | No |
| Read an article as clean plain text (internal blocks expanded) or HTML for editing | No |
| Filter by category, published/draft, visibility, dates, language | No |
| Published articles not updated in N days, oldest first | No |
| Category tree with article counts, or one category's articles | No |
| Who created, edited or published what | No |
| Account settings | No |
| New article, saved as a draft by default; accepts HTML or plain text | Yes |
| Change only the fields you pass | Yes |
| New category or subcategory | Yes |
| Delete an article | Destructive |
Design notes:
Article bodies come back as readable text instead of raw HTML, which keeps answers grounded and saves tokens. Ask for
format: "html"when editing.New articles are unpublished drafts unless you explicitly set
published: true, so an agent can't put content in front of customers by accident.Lists return compact rows (id, title, visibility, views, last updated) so an agent can scan hundreds of articles in one call.
All tools carry MCP annotations (
readOnlyHint,destructiveHint), so clients can ask before writing.
Setup
In Helpjuice, go to Settings → API Credentials, enable token-based API access and copy your API key.
Build it:
git clone https://github.com/OfirOhan/helpjuice-mcp.git
cd helpjuice-mcp && npm install && npm run buildClaude Desktop
Add this to claude_desktop_config.json:
{
"mcpServers": {
"helpjuice": {
"command": "node",
"args": ["/absolute/path/to/helpjuice-mcp/dist/index.js"],
"env": { "HELPJUICE_ACCOUNT": "yourcompany", "HELPJUICE_API_KEY": "your-api-key" }
}
}
}Claude Code / Cursor / other MCP clients
claude mcp add helpjuice -e HELPJUICE_ACCOUNT=yourcompany -e HELPJUICE_API_KEY=your-api-key -- node /path/to/helpjuice-mcp/dist/index.jsFor Cursor and other clients, use the same command with the variables below in the environment.
Variable | Default | Notes |
| (required) | Your Helpjuice API key |
| (required) | Your subdomain, e.g. |
|
| Override for testing |
Development
npm install
npm test # builds, runs unit tests and an end-to-end MCP stdio test against a fake Helpjuice APIThe tests run on Node 20, 22 and 24 in CI.
Author
Built by Ofir Ohana, an AI agents engineer. Issues and PRs are welcome.
License
MIT
Available Tools
12 toolscreate_articleCreate articleA
Create a new article. Saved as an unpublished draft unless published=true. body accepts HTML or plain text (blank lines become paragraphs).
Inputs: body (required), name (required), codename, language, +3 more.
create_categoryCreate categoryC
Create a knowledge base category, optionally nested under a parent.
Inputs: name (required), parentId, description, accessibility.
delete_articleDelete articleADestructive
Delete an article from the knowledge base. Confirm with the user first; consider unpublishing instead.
Inputs: articleId (required).
find_stale_articlesFind stale articlesARead-only
Find published articles that have not been updated in N days (default 180), oldest first. Useful for content audits before a release.
Inputs: days, limit, categoryId.
get_accountAccount settingsARead-only
Get the knowledge base account settings (name, subdomain, internal KB flag, contact email).
Inputs: no named parameters.
get_articleRead articleARead-only
Read one article. Returns the body as clean plain text by default (internal blocks expanded), or as HTML for editing.
Inputs: articleId (required), format, language.
get_categoryGet categoryARead-only
Get one category with its published and draft articles.
Inputs: categoryId (required).
list_articlesList articlesBRead-only
List articles with filters (category, published, visibility, created/updated dates, language). Returns compact rows.
Inputs: page, limit, language, published, +4 more.
list_categoriesList categoriesARead-only
List knowledge base categories (id, name, parent, visibility, article counts).
Inputs: language.
list_recent_activityRecent activityARead-only
Recent knowledge base activity (who created, edited or published what), newest first.
Inputs: page, limit, actionType, trackableType.
search_knowledge_baseSearch knowledge baseARead-only
Full-text search across the Helpjuice knowledge base. Returns matching articles with a short text sample. Use get_article to read the full answer before quoting it.
Inputs: query (required), categoryId.
update_articleUpdate articleAIdempotent
Update an article's title, body, summary, categories or published state. Only the fields you pass are changed. Read the article first with get_article(format='html') when editing the body.
Inputs: articleId (required), body, name, published, +2 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 12 tools
Most tools target clearly distinct actions and resources, with article CRUD, category listing/creation, and account/activity reads separated well. The main possible confusion is between search_knowledge_base, list_articles, and find_stale_articles, though their descriptions clarify the intended use cases.
All tool names follow a consistent snake_case verb_noun pattern such as search_knowledge_base, get_article, list_articles, create_article, and delete_article. The convention is predictable throughout the set.
Twelve tools is well-scoped for a knowledge base server, covering article and category management plus search and activity inspection. No tool appears redundant or gratuitous at the surface level.
Article lifecycle coverage is strong with search, read, list, create, update, delete, and stale-content auditing. Category management is less complete because categories can be listed, fetched, and created but not updated or deleted.
Related MCP Connectors
Connect your team's living knowledge base — docs, data, issues, CRM — to Claude and ChatGPT.
Connect Claude, ChatGPT or any AI assistant to Zendesk tickets, CSAT, macros and Help Center.
Connect AI assistants to Plain customer support data and workflows.
Work your HelpRabbit help desk: tickets, replies, customers, the AI agent and its knowledge.
Related MCP Servers
- FlicenseAqualityDmaintenanceEnables natural language query and ingestion of wiki knowledge bases built from documents, powered by Claude Agent SDK skills.2-
- FlicenseNot gradedqualityDmaintenanceEnables Claude Desktop to query a local Zendesk Help Center knowledge graph for semantic search, article retrieval, content gap analysis, and editorial intelligence tasks.3-
- FlicenseNot gradedqualityDmaintenanceGives Claude access to your Help Scout knowledge base for listing collections, articles, and searching content.-
- AlicenseAqualityDmaintenanceEnables AI agents to browse, search, and fetch documentation from any public Zendesk Help Center via a read-only API.9MIT