Your AI. Your key. Your server.
FreeITSM has AI woven through the modules you use every day — answering questions, summarising, drafting, tidying up. Every bit of it is opt-in, bring-your-own-key, and configured per feature, with a choice of Anthropic, OpenAI, or hundreds of models through OpenRouter.
What the AI actually does
No vague "powered by AI" badge. Each feature does one concrete, useful thing in the module where you already work — and every one is switched off until you give it a key.
Ask AI
Ask a question in plain English and get an answer drawn from your own knowledge articles, with the sources cited. Optional semantic (vector) search finds the right articles even when the words don't match.
Read the deep diveConfiguration item summaries
A plain-English synthesis of each configuration item — what it is, what it depends on, and recent ticket activity around it — so a CMDB record earns the next visit instead of becoming a write-only graveyard.
Read the deep diveAI co-author
Describe the automation you want in a sentence and the co-author scaffolds a first draft on the canvas — triggers, conditions and actions — ready for you to refine rather than build from a blank page.
Read the deep diveAI Assist streams live
The form builder isn't slow at building forms — it's slow at starting them. Describe a form in a paragraph and watch a working draft appear, field by field, ready to tweak and save.
Read the deep diveReply cleanup streams live
Type a one-line shorthand reply and a Cleanup button turns it into a proper email — greeting, context, sign-off — with strict guards so the AI tidies what you wrote rather than inventing facts.
Read the deep diveRFP Builder streams live
Turn five messy department requirement docs into one polished, de-duplicated RFP. The AI does the heavy lifting of merging and structuring; you stay in control of the result.
Read the deep divePick a provider. Pick a model. Paste a key.
There's no single AI switch for the whole app. Each feature is configured on its own module's settings page, with its own provider, model and key — so you choose exactly what powers what.
Encrypted at rest
Keys are stored AES-256-GCM encrypted in your own database, shown only as a mask, and never sent back to the browser.
Per feature
Run a cheap, fast model for reply cleanup and a stronger one for workflow authoring. Each feature is independent.
Test before you trust
A Test button on every panel confirms the provider and key work before you rely on the feature in anger.
The shared provider layer and the OpenRouter integration are recent additions — if you want the full architecture story, the deep dive walks through how one drop-in building block gives every feature the same three-provider choice. Read “One key, hundreds of models” →
Setting it up
Configuring an AI feature takes about a minute. The panel is the same everywhere, so once you've done one you've done them all.
Get a key from your provider
Create an API key at Anthropic, OpenAI, or OpenRouter. For the widest model choice from a single account, OpenRouter is the easy pick. The API setup guide has step-by-step instructions.
Set a spending cap first
Before you paste a key anywhere, set a hard spending limit on the provider account. AI calls cost money per token and those costs are yours — a cap is your safety net.
Open the feature's settings
Go to the module that owns the feature — for example Knowledge → Settings, or Tickets → Settings → Reply cleanup — and find its AI panel.
Choose provider, model and key
Pick Anthropic, OpenAI or OpenRouter. With OpenRouter, the model field becomes a searchable list with live pricing; with the others, type or pick a model. Paste your key and leave SSL verification on.
Test, then save
Hit Test to confirm the connection, then Save. The key is encrypted immediately. Re-saving the model later won't disturb the key — you only change it by typing a new one.
What leaves your server, and when
FreeITSM is self-hosted, and the AI features are built to respect that.
The plain version
- Nothing happens until you opt in. No AI feature does anything — or sends anything anywhere — until you configure it with a key.
- Data only goes to the provider you chose. When a feature runs, the relevant prompt is sent to your selected provider to get an answer, and nowhere else.
- OpenRouter routing is disclosed. If you choose OpenRouter, prompts are routed through OpenRouter's service to reach the model — the settings panel says so plainly. Prefer a direct line? Pick Anthropic or OpenAI.
- Your keys stay yours. Encrypted at rest on your own database, never exposed to the browser, never sent anywhere except as the provider's own auth header.
- No FreeITSM middleman. There is no FreeITSM cloud in the loop, no telemetry on your prompts, and no per-seat AI tax. It's your account talking to your provider.
Bring your own key. Keep your own data.
FreeITSM is free, open source and self-hostable. The AI is there when you want it, off when you don't, and always powered by your account on your terms.