Guru MCP

Access Guru knowledge base via MCP - ask AI questions, search documents, create drafts, and update cards. Connects to all your Guru sources including Slack, Drive, Confluence, and SharePoint.

Security Scan

The skill's behavior (calling Guru's MCP via the mcporter CLI using a GURU_API_TOKEN) is coherent with its description, but there are inconsistencies between the published metadata and the runtime instructions and it asks you to write credentials/config files to specific paths — review before installing.

Purpose & Capability

The skill's stated purpose (access Guru via MCP to search, ask questions, create/update cards) matches the SKILL.md instructions: it expects an MCP endpoint, a GURU_API_TOKEN, and the mcporter client to call the MCP API. However, the registry metadata provided with the skill lists no required binaries or env vars, while the SKILL.md metadata and instructions explicitly require the mcporter binary and a GURU_API_TOKEN — this mismatch is an incoherence to investigate.

Instruction Scope

Instructions are specific and constrained to the stated task: create a Guru API token, set GURU_API_TOKEN in ~/.clawdbot/.env, add a guru entry to config/mcporter.json, and invoke mcporter calls to MCP endpoints. These steps involve reading/writing configuration files in your home directory and calling an external API (mcp.api.getguru.com). The instructions do not ask for unrelated files/credentials, but they do instruct creating/modifying files at specific paths which is outside the skill bundle and should be done only if you trust the source.

Install Mechanism

This is an instruction-only skill with no install spec and no code to write to disk. That lowers risk. It does rely on an external CLI (mcporter) being present but does not provide an installer — the absence of an install spec is consistent with an instruction-only integration, though you should obtain mcporter from an official source.

Credentials

The only sensitive credential the SKILL.md requires is a single GURU_API_TOKEN (in the form email:token), which is proportionate for a Guru integration. That token is used in an Authorization header for the MCP server. However, the top-level registry metadata claimed no required env vars while the embedded SKILL.md requires GURU_API_TOKEN — this inconsistency should be resolved. Also note that queries are logged in Guru's AI Agent Center (per the docs), so content sent to the skill will be recorded by Guru.

Persistence & Privilege

The skill does not request 'always: true' or other elevated platform privileges. It will operate only when invoked. It does instruct you to write configuration files (~/ .clawdbot/.env and config/mcporter.json), which is normal for configuring a CLI client but is a permanent change to your environment that you should be willing to make before proceeding.

What to consider before installing

Before installing or using this skill: (1) Confirm the source — the registry shows 'Source: unknown' but the homepage is getguru.com; prefer installing only if you trust that origin. (2) Resolve the metadata mismatch: the SKILL.md requires the mcporter binary and a GURU_API_TOKEN even though the registry metadata lists none — verify that requirement and obtain mcporter from an official source. (3) Create a scoped/least-privilege Guru API token, understand its permissions and retention, and consider using a token you can revoke. (4) Be aware the skill instructs writing credentials to ~/.clawdbot/.env and editing config/mcporter.json — these are persistent changes; review those files and file permissions. (5) Remember that queries are logged in Guru's AI Agent Center, so do not send sensitive secrets or private data unless you accept that logging. If you cannot verify the origin or the mcporter client, treat this skill as untrusted.

Access your Guru knowledge base via the official MCP server. Ask AI-powered questions, search documents, create drafts, and update cards.

Features

• AI-Powered Answers — Get comprehensive answers from Knowledge Agents
• Document Search — Find cards and content across your knowledge base
• Create Drafts — Generate new card drafts from AI tools
• Update Cards — Modify existing cards directly
• Connected Sources — Access Salesforce, Slack, Google Drive, Confluence, SharePoint through Guru
• Permission-Aware — Respects all existing Guru permissions
• Analytics — All queries logged in AI Agent Center

Setup

1. Get API Token

1. Go to Guru Admin → API Tokens
2. Create a new token
3. Note your email and token

2. Configure Environment

Add to ~/.clawdbot/.env:

GURU_API_TOKEN=[e-mail]:your-api-token

3. Configure mcporter

Add to config/mcporter.json:

{
  "mcpServers": {
    "guru": {
      "baseUrl": "https://mcp.api.getguru.com/mcp",
      "headers": {
        "Authorization": "Bearer ${GURU_API_TOKEN}"
      }
    }
  }
}

4. Verify

mcporter list guru

Available Tools

guru_list_knowledge_agents

List all Knowledge Agents in your workspace. Always call this first to get agent IDs for other tools.

mcporter call 'guru.guru_list_knowledge_agents()'

Returns:

[
  {"id": "08de66e8-...", "name": "Guru"},
  {"id": "abc123...", "name": "Engineering Docs"}
]

guru_answer_generation

Get AI-powered answers from a Knowledge Agent. Best for specific questions like "What is X?" or "How do I Y?".

mcporter call 'guru.guru_answer_generation(
  agentId: "YOUR_AGENT_ID",
  question: "How do I submit expenses?"
)'

Optional filters:

• collectionIds — Limit to specific collections
• sourceIds — Limit to specific sources

Returns comprehensive answer with sources.

guru_search_documents

Find documents, cards, and sources. Best for browsing content like "find docs on X" or "do we have cards about Y?".

mcporter call 'guru.guru_search_documents(
  agentId: "YOUR_AGENT_ID",
  query: "onboarding process"
)'

Returns list of matching documents with snippets.

guru_get_card_by_id

Get full card content in HTML format.

mcporter call 'guru.guru_get_card_by_id(id: "CARD_ID")'

Returns card ID, title, and HTML content.

guru_create_draft

Create a new card draft.

mcporter call 'guru.guru_create_draft(
  title: "New Process Guide",
  content: "<h2>Overview</h2><p>This guide covers...</p>"
)'

Returns draft ID and URL.

guru_update_card

Update an existing card. First retrieve current content with guru_get_card_by_id, then modify.

mcporter call 'guru.guru_update_card(
  cardId: "CARD_ID",
  title: "Updated Title",
  content: "<p>Updated HTML content...</p>"
)'

Important: Preserve HTML structure when updating. Insert/replace content within existing DOM hierarchy.

Usage Patterns

Ask a Question

# 1. Get agent ID
mcporter call 'guru.guru_list_knowledge_agents()'

# 2. Ask question
mcporter call 'guru.guru_answer_generation(
  agentId: "08de66e8-...",
  question: "What is the PTO policy?"
)'

Find and Read a Card

# 1. Search for cards
mcporter call 'guru.guru_search_documents(
  agentId: "08de66e8-...",
  query: "expense report"
)'

# 2. Get full content
mcporter call 'guru.guru_get_card_by_id(id: "CARD_ID_FROM_SEARCH")'

Create New Documentation

mcporter call 'guru.guru_create_draft(
  title: "API Authentication Guide",
  content: "<h2>Overview</h2><p>This guide explains how to authenticate with our API.</p><h2>Steps</h2><ol><li>Generate API key</li><li>Add to headers</li></ol>"
)'

Choosing the Right Tool

Use Case	Tool
"What is X?" / "How do I Y?"	guru_answer_generation
"Find docs about X"	guru_search_documents
"Show me card XYZ"	guru_get_card_by_id
"Create a new guide for X"	guru_create_draft
"Update this card with..."	guru_update_card

Token Format

The GURU_API_TOKEN must be in format email:token:

[e-mail]:a1b2c3d4-e5f6-7890-abcd-ef1234567890

Notes

• Questions appear in Guru's AI Agent Center analytics
• All permissions enforced (users only see what they have access to)
• Knowledge Agents can be domain-specific — choose the right one for your question
• Card content is HTML — preserve structure when updating

Resources

• Guru MCP Documentation
• Guru API Reference
• AI Agent Center
• MCP Feedback