AI on your terms, metered by nobody
Get AI-drafted replies that quote your own knowledge base, billed by nobody but the provider you already trust. Or switch AI off entirely. QuerySolv never meters a resolution.
- Hooks for Claude, ChatGPT and Grok, any HTTP endpoint, or a local command. Or no AI at all.
- Drafts can be grounded in your knowledge base and in similar resolved tickets.
- A draft fills the composer. An agent reads it and decides; nothing is sent automatically.
- QuerySolv meters nothing. A hosted provider sees the ticket; the local-command hook can keep it on the server.
Bring your own provider
Use the provider your organization has approved, a model you host yourself, or nothing at all. No AI is forced on you.
Drafts grounded in your own answers
Grounding is an option you switch on for each hook. With it on, a draft request also carries up to three of the best-matching knowledge-base articles and the closing agent reply from up to three similar resolved tickets in the same department.
Both are found by the built-in search and re-checked against live records, so a draft sounds like your team and cites your answers instead of generic text.
A copilot, not an autopilot
The agent decides
A reply draft fills the composer. It is never stored and never sent automatically. An agent reads it and decides.
Bounded requests
Requests have a one-minute limit and responses a 256 KB cap.
Keys stay secret
Provider keys are encrypted at rest and never written to the log.
An optional first answer in live chat
You can let a bot answer first in live chat. It can only say what your knowledge base already says.
- A visitor asksThe bot answers first, before an agent is involved.
- It answers from your articlesIt uses only published, visible knowledge-base articles, so it can only say what your knowledge base already says.
- A human takes overAfter four answers, on request, or when it cannot help, the conversation is handed to an agent.
The chat bot needs an enabled Claude, ChatGPT or Grok hook. It never uses a raw or local-command hook. It is switched on by editing the configuration file and restarting the service, one of a few file-only settings.
The live chat widget. With the bot enabled, the first answers come from your knowledge base.
AI customer service software, without the platform
An "AI-powered customer service platform" is usually a cloud suite with the model bundled in, metered per resolution and trained on nobody knows what. QuerySolv does the three jobs that actually save an agent's time, with a model you chose, and leaves the rest alone.
Drafts the reply
For any ticket, from any channel, grounded in your articles and your past answers. The agent edits and sends, or discards.
Answers first in chat
The optional bot gives the first answers in live chat from your published knowledge base, and hands over to a person after four answers, on request, or when it cannot help.
Automates nothing you did not ask for
No ticket is closed, replied to or reclassified by a model on its own. Customer service automation here means a faster agent, not an absent one.
What it costs
Whatever you agree with the provider you pick. QuerySolv adds nothing on top.
Elsewhere
AI billing has moved from per seat to per outcome: a fee for each automated resolution, or packs of sessions that run out. That makes a support budget hard to predict.
Here
The only variable is your own provider's bill, and with a local model there may be no bill at all.
What leaves the building
You should know this before you enable anything. The manual states all of it plainly.
A hook that calls a hosted provider sends
- That ticket's conversation, including internal notes
- The requester's name
- With grounding on, the articles and closing replies described above
The local-command hook keeps it inside
If that is not acceptable, use the local-command hook. It runs a program on the server and passes the context on standard input, so an on-premises model keeps the ticket history inside the building.
AI assistance is part of the Professional and High Availability editions. See compare editions.
Try it with your own provider
Download the installer, start the service, add a key to one of the seeded hooks and ask for a draft on a test ticket. It runs for 30 days with every Professional edition feature and unlimited agents.
Download Installer with Free Trial