Manual
AnyLLM User Manual
Welcome. This manual describes how to use AnyLLM in Jira and Confluence — from the perspective of everyday work and from the perspective of administration. Use the index below to find the page you need.
What is AnyLLM?
AnyLLM brings an AI chat directly into Jira and Confluence. You ask your question where you already work — in the issue, on the page, on the dashboard — and receive an answer that knows the context.
- Ask without switching tools — no second tab, no copying content into another chat window.
- Work with your own content — the AI can search and read issues and pages, and only those you are allowed to see yourself.
- Suggestions instead of unilateral action — if something is to be created or changed (issue, comment, page), the AI shows a suggestion card; it is carried out only after your click.
- Your organisation's own AI service — your administration connects the language models it has chosen itself. The API keys stay encrypted at Atlassian.
Who this manual is for
| Role | Who that is | What this role can do | Start here |
|---|---|---|---|
| Standard user | anyone with access to Jira or Confluence | chat, create their own agents and skills, manage history, use quick actions | Getting Started |
| Product administrator | Jira administrator or Confluence administrator | everything from the user perspective plus setup: connections, endpoint approvals (Jira administrators only), agents, templates, modules, general settings | Administration & Configuration |
There are no other roles. Who may see which content is decided solely by your existing Jira and Confluence permissions — the app grants no permissions of its own. Details: Permissions, Roles and Data Visibility.
Prerequisites
| Prerequisite | Who takes care of it | What happens without it |
|---|---|---|
| The app is installed in Jira (required); Confluence is added to the same installation afterwards — a Confluence-only installation is not possible | Product administration | AnyLLM appears nowhere (or not in Confluence) |
| At least one connection to an AI service with a saved API key | Product administration | Notice Not set up yet, input field locked |
| The endpoint address is approved in each product (once in Jira, once in Confluence) | Jira administration | Requests are blocked; error message when sending |
| For a paid Marketplace installation: active subscription | Site administration | Notice banner, no new answers |
| The respective surface is enabled (e.g. issue panel) | Product administration | Notice This feature has been disabled by your administrator. |
Quick start
For users (2 minutes)
- In Jira or Confluence open Apps → AnyLLM Chat.
- Click into the field Message AnyLLM … and type your question.
- Press Enter. The answer appears while it is being generated.
Continue with Getting Started and Chatting with the AI.
For administrators (5 steps)
- Open the settings — Jira: Settings → Apps → AnyLLM Chat — Settings; Confluence: Settings → AnyLLM Chat — Settings.
- Create a connection under Connections (name, endpoint address, model names, API key) and save.
- Confirm the endpoint approval under Endpoint approvals — once in Jira and once in Confluence.
- Under General set the interface language, the default model and the AI access to Jira/Confluence.
- Ask a test question in the chat.
Details and the safe order for the initial setup: Administration & Configuration.
All pages
| Page | What you find there |
|---|---|
| Getting Started | where AnyLLM appears in Jira and Confluence, the first screen, the interface tour, your first question |
| Chatting with the AI | asking questions, choosing agent and model, regenerate, stop, export, chips, remembered choices |
| Prompts, Skills, Memory and Model Parameters (＋ Menu) | the ＋ menu: prompts and skills, memory, model parameters, page context, temporary chat |
| Creating Your Own Agents and Skills | own agents (create, share, use, edit, delete), own skills |
| Letting the AI Search, Read and Create in Jira and Confluence | lookups, suggestion cards, rich text, web search, image generation |
| Managing Chat History: Pin, Fork, Archive, Delete | opening, limits, pinning, forking, deleting, archiving to Confluence |
| AnyLLM in Confluence — Chat Bar, Quick Actions, Page Assistant and Macro | all Confluence surfaces |
| AnyLLM in Jira — Issue Panel, Sidebar Panel, Issue Navigator and Command Palette | all Jira surfaces |
| Dashboard Gadget (AI Report) | a saved question answered on the dashboard |
| Permissions, Roles and Data Visibility | roles, principles, who may do what, who sees what, what goes where |
| Administration & Configuration | index of the administration section and the safe order for the initial setup |
| Opening the Settings and Initial Setup | opening the settings in both products, navigation, changes during operation |
| Setting Up LLM Connections | create, test, change key, delete key, remove connection |
| Managing Endpoint Approvals | per-product approvals: approve, check, replace, revoke, re-approve |
| Providing Agents, Prompts and Skills for Everyone | site agents, prompts and skills |
| General Settings: Language, Default Model, AI Access, Rate Limit, Site Knowledge | the General page |
| Enabling Web Search and Image Generation | search services and image generation |
| Modules: Switching Surfaces On and Off | all module toggles and their effects |
| Usage Statistics, Import/Export and Backup | usage page, export, import |
| Licence, Data Retention and Automatic Deletion | subscription, data storage, deletion paths, weekly cleanup |
| Troubleshooting | messages verbatim, cause, solution |
| FAQ | frequently asked questions for users and administrators |
| Glossary | terms A–Z and the English/German label comparison |
About the interface language
This manual quotes the English labels — the state in which the app is shipped. The administration can switch the interface under General → Interface language to German (Deutsch) or Follow each user’s Atlassian language; the German edition of this manual quotes the German labels. Menu entries in Jira and Confluence (e.g. AnyLLM Chat) always follow the Atlassian language of the individual person. Messages from the server appear in English even in a German interface; the only exception is the notice about an endpoint that is not yet approved, which is shown in the interface language. The sidebar labels Chat, Agents and Skills contain no emoji — the icons next to them are part of the design; some entries of the ＋ menu do carry an emoji prefix (e.g. 🧠 Edit memory), which this manual quotes verbatim.
Known limitations
These points are the case in the current version and are not malfunctions:
| Limitation | What that means | Details |
|---|---|---|
| History, memory, agents and skills span Jira and Confluence | AnyLLM is one installation for both products. A conversation started in Confluence is also listed in Jira — only for the same person. | Permissions |
| 50 conversations are kept | Older, unpinned conversations are removed automatically. | History |
| The AI only sees the last 16 messages | Only the most recent 16 messages (about 32,000 characters) of a conversation go to the AI; earlier turns stay in the history but are no longer remembered. | Chat |
| Stop only hides the answer | The running answer disappears from the screen, but the request continues on the server and the full answer may still land in the saved history. | Chat |
| Menu entries of disabled surfaces stay visible | A switched-off surface still has its menu entry; it only opens the notice This feature has been disabled by your administrator. Exception: the chat bar in Confluence hides itself completely. | Modules |
| The chat bar reacts with up to 10 minutes delay — in both directions | The chat bar's status is cached in the browser for ten minutes: after switching it off it may stay visible for up to ten minutes, and after switching it on (or completing the setup) it may stay hidden for up to ten minutes. | Confluence surfaces |
| No full-text search across conversations | The history cannot be searched. Archive conversations to Confluence if you want to search them. | History |
| No per-person usage | The usage statistics show requests and tokens per model for the whole site — without content and without attribution to individuals. | Usage statistics |
State of this manual
- Describes AnyLLM as of 25 August 2026.
- Produced from the source code, the interface texts and the module and permission definitions of the app.
- Screenshots are not yet included; the places where they will be added are prepared in the pages.
Related topics
Documentation baseline: app version 0.2.0 · 2026-08-30