Skip to content
🛟

Support Knowledge Base

A shape for the support knowledge that ships once and pays back forever. Issues, resolutions, escalation paths, known limitations. Claude and ChatGPT answer support questions from real past resolutions.

Requires an AI connected to your Hjarni account via MCP.

https://hjarni.com/templates/support-knowledge-base

Copy this URL and paste it into Claude or ChatGPT to install the template.

How to use

  1. 1 Share this page. Paste this URL into Claude or ChatGPT. Your AI reads the template definition and installs it.
  2. 2 Folders, tags, and instructions appear. Your AI creates the full structure in your Hjarni account, ready to use.
  3. 3 Start adding notes. The AI instructions guide your AI on where to put things and how to organize them.

Every solved ticket is a future answer. Claude and ChatGPT reply from real resolutions, not generic templates.

A small support team without an internal knowledge base solves the same ticket five times. This template gives those resolutions a shape your AI can read the next time the issue comes in.

A resolution note format

Every resolution follows the same shape, ending with a customer-ready reply.

# Resolution: <Short outcome>

Symptom:
Root cause:
Steps that worked:
Customer-facing reply template:
Source ticket:

Issues, Resolutions, and Known Limitations stay separate

The Issues folder collects symptoms. The Resolutions folder collects fixes. The Known Limitations folder collects honest answers about what the product cannot do. Keeping them apart stops the AI from quietly promoting a workaround into a roadmap promise.

A workflow that earns the template's keep

  1. When a ticket is resolved, ask Claude to extract a Resolution note from the conversation.
  2. Tag the matching Issue note with the new ticket reference. Do not create a duplicate Issue.
  3. When the same question recurs, ask Claude to draft the reply using the Resolution's customer-facing template.
  4. When a limitation comes up in a ticket, point at the Known Limitations note instead of inventing a workaround.

A real example

A new ticket asks why the first page load is slow. Your support teammate asks Claude, "How do we usually answer this?" Claude reads the Issues folder, finds the matching note, pulls the Resolution, and returns the customer-facing reply template. Teammate edits one line for tone, sends it. Five minutes instead of an hour.

Common questions

Common questions

Is this a replacement for Zendesk or Intercom?

No. Your help desk owns the inbox. Hjarni holds the resolutions the help desk does not capture, so they survive across tickets and teammates.

Will the AI invent customer replies?

Folder instructions tell each AI to draft from the Resolution's customer-facing template, not from training data. If the resolution does not exist, the AI says so.

Can my support team share this?

Yes. Install in a team space so every teammate's Claude or ChatGPT reads from the same resolutions and limitations.

What about Known Limitations? Does the AI accidentally promise the workaround as a fix?

Folder instructions tell each AI to treat Known Limitations as binding. The workaround is offered, the limitation is not papered over.

Related pages

Structure

Tags

issue resolution escalation known-limitation faq

Folders

Support Knowledge Base
Your accumulated support memory. Every solved ticket is a future answer.
Has AI instructions
5 starter notes
Issues
One note per recurring issue. Symptoms, affected users, frequency.
Has AI instructions
Resolutions
How you actually fixed each issue. The reusable answer.
Has AI instructions
Escalation Paths
Who to ping when an issue jumps tiers. Operational steps.
Has AI instructions
Known Limitations
What the product cannot do today. Honest answers, not roadmap promises.
Has AI instructions
Public FAQ
Customer-facing answers ready to send or post.
Has AI instructions

For your AI

Share this page with your AI. It reads the definition below, creates the folders, instructions, tags, and starter notes in your account.

Show template definition

Install steps for AI agents:

  1. Check existing tags with tags-list. Only create missing ones with tags-create.
  2. Create containers top-down using containers-create, noting the returned IDs. Use parent_id to build the hierarchy.
  3. For each container with llm_instructions, call instructions-update with level: "container" and the container's ID.
  4. Create any seed notes using notes-create, placing them in the correct container by ID. Use container_path to resolve which container.
  5. Discuss any customizations with the user before or after installing.
---
name: Support Knowledge Base
description: 'A shape for the support knowledge that ships once and pays back forever.
  Issues, resolutions, escalation paths, known limitations. Claude and ChatGPT answer
  support questions from real past resolutions.

  '
tags:
- issue
- resolution
- escalation
- known-limitation
- faq
containers:
- name: Support Knowledge Base
  description: Your accumulated support memory. Every solved ticket is a future answer.
  llm_instructions: |
    This is a support knowledge base. The goal is to make every past resolution useful for the next ticket and to keep the team and the AI aligned on what is true.
    - When asked about a user-facing issue, search Issues and Resolutions first. Cite the source resolution note.
    - Never invent a fix, a workaround, or a product limitation. If the answer is not in this folder, say so plainly and ask before drafting a customer reply.
    - When the user resolves a new ticket in conversation, suggest writing a Resolution note so the answer becomes searchable next time.
    - Treat Known Limitations as binding. If a customer asks for something on that list, do not promise it.
    - The Escalation Paths folder is operational. Follow it exactly. Do not improvise an escalation route.
  children:
  - name: Issues
    description: One note per recurring issue. Symptoms, affected users, frequency.
    llm_instructions: |
      Use this folder for the patterns customers report.
      - One note per issue. Title format: "Issue: <Short symptom>".
      - Include: Symptoms, Who hits it, How often, Linked resolutions.
      - When a new ticket matches an existing issue, link the new ticket details inside the matching note rather than creating a duplicate.
      - Tag every note with "issue".
  - name: Resolutions
    description: How you actually fixed each issue. The reusable answer.
    llm_instructions: |
      Use this folder for the answer that actually solved a customer problem.
      - One resolution per note. Title format: "Resolution: <Short outcome>".
      - Include: Symptom, Root cause, Steps that worked, Customer-facing reply template, Source ticket.
      - When asked "how do I handle X", return the matching resolution and quote the customer-facing reply template.
      - Tag every note with "resolution".
  - name: Escalation Paths
    description: Who to ping when an issue jumps tiers. Operational steps.
    llm_instructions: |
      Use this folder for the escalation routes.
      - One note per path. Title format: "Escalation: <Issue type>".
      - Steps should be numbered and executable. Names and channels should be current.
      - When asked who to ping, return the matching path. Do not invent an alternate route.
      - Tag every note with "escalation".
  - name: Known Limitations
    description: What the product cannot do today. Honest answers, not roadmap promises.
    llm_instructions: |
      Use this folder for things the product genuinely cannot do.
      - One note per limitation. Title is a short sentence in the negative ("Cannot import from X").
      - Include: What the limitation is, Why it exists, A workaround if any, Whether it is on the roadmap.
      - When a customer asks for something on this list, do not promise. Quote the workaround if there is one.
      - Tag every note with "known-limitation".
  - name: Public FAQ
    description: Customer-facing answers ready to send or post.
    llm_instructions: |
      Use this folder for polished, customer-ready answers.
      - One FAQ per note. Title is the question, exactly as a customer would ask it.
      - Each note has two parts: Short answer (the reply) and Source (the internal resolution or issue note).
      - When drafting a reply, prefer the FAQ wording over rewriting from scratch.
      - Tag every note with "faq".
  notes:
  - title: 'Issue: Slow first page load after sign-in'
    body: |
      A starter issue note. Replace with a real recurring issue.

      ## Symptoms
      Users report a 6 to 10 second blank screen after signing in for the first time.

      ## Who hits it
      Mostly new users on smaller workspaces. Repeat visits are fast.

      ## How often
      Roughly two tickets a week since launch.

      ## Linked resolutions
      - [[Resolution: Warm the workspace cache on sign-up]]

      This is a starter note. Replace it with a real issue.
    tags:
    - issue
    container_path: Support Knowledge Base > Issues
  - title: 'Resolution: Warm the workspace cache on sign-up'
    body: |
      A starter resolution note. Replace with a real one.

      ## Symptom
      Slow first page load after sign-in.

      ## Root cause
      The dashboard query loads the entire workspace tree. On first visit, nothing is cached.

      ## Steps that worked
      1. Enqueue the cache warmup job on the sign-up confirmation.
      2. Show the spinner state on the dashboard for first-time users only.
      3. Send a confirmation email letting the user know the workspace is ready.

      ## Customer-facing reply template
      > Thanks for flagging. The first sign-in is slower because we warm up the workspace in the background. Subsequent visits load instantly. We are working on shortening that first visit too.

      ## Source ticket
      Replace with a real ticket reference.

      This is a starter note. Replace it with a real resolution.
    tags:
    - resolution
    container_path: Support Knowledge Base > Resolutions
  - title: 'Escalation: Data loss reports'
    body: |
      A starter escalation note. Replace with the real path.

      ## When to use
      Any report where a user says notes, folders, or attachments have disappeared.

      ## Steps
      1. Acknowledge the report within one hour.
      2. Tag the ticket with "data-loss" and assign to the on-call engineer.
      3. Pull the user's recent activity log and confirm whether the records are recoverable.
      4. Reply with a status update within three hours, regardless of progress.
      5. If recovery requires a backup restore, follow the Backup Restore runbook (link to be added once Developer Project Memory is connected).

      This is a starter note. Replace it with a real escalation path.
    tags:
    - escalation
    container_path: Support Knowledge Base > Escalation Paths
  - title: Cannot import from Notion at this time
    body: |
      A starter limitation note. Replace with a real one.

      ## What the limitation is
      We do not currently offer a direct Notion importer.

      ## Why it exists
      Notion's export format is HTML with custom blocks. We have not built a parser yet.

      ## Workaround
      Users can export Notion pages to Markdown, then drop the files into Hjarni manually.

      ## Roadmap status
      Not currently scheduled.

      This is a starter note. Replace it with a real limitation.
    tags:
    - known-limitation
    container_path: Support Knowledge Base > Known Limitations
  - title: Why is the first page load slow?
    body: |
      A starter FAQ note. Replace with a real customer-ready answer.

      ## Short answer
      The first sign-in is slower because we warm up your workspace in the background. After that first visit, everything loads instantly. We are working on shortening that first visit too.

      ## Source
      [[Resolution: Warm the workspace cache on sign-up]]

      This is a starter note. Replace it with a real FAQ.
    tags:
    - faq
    container_path: Support Knowledge Base > Public FAQ

Write once. You both remember.

Free to start. No credit card required.

Give your AI a memory

Works with Claude and ChatGPT today. Gemini coming soon.