Jan
Open-source desktop app for running local models, good for individuals.
Msty is one of the nicest ways to run AI privately on your own computer, and it now sells a teams tier. If your firm needs everyone working from the same files, with a record of what was asked and answered, a team workspace on your own servers may fit better.
Msty is a private AI app that runs on each person's computer, with local and cloud models in one place. Its Enterprise and Teams tier adds SSO, user management and admin audit logs, while chats stay on each device.
Locul is one shared private workspace for the whole firm, running on servers in the UK, the EU or your building, with a log of every question and answer. Choose Msty for personal use; consider Locul when the team needs one brain and one record.
Msty is well made and privacy-minded. If you work alone or in a very small team, it is hard to beat.
None of these are faults in Msty. They come from its design: a great app on each computer, rather than one service for the firm.
Chats live on each device and are not synced, so the answer Tom got is not the one Priya can find.
Msty's audit log covers admin actions such as users and teams, not conversations or which documents were used.
Shared items need S3-compatible storage that your firm provides and looks after.
We found no approval workflow for emails or posts drafted by the AI.
Checked against Msty's own website and documentation in October 2026. Products change quickly, so check anything that matters to you on their site too.
| Msty | Locul | |
|---|---|---|
| How it runs | A desktop app on each computer, plus a web version that connects to it | A browser app for everyone, served from one private deployment |
| Where models run | On each user's machine, or a cloud provider | Dedicated servers in the UK, the EU or your building |
| Shared team knowledge | Shared Knowledge Stacks via S3 storage you supply | One team brain over your drives and wiki |
| Permissions | Teams, access controls and RBAC on the teams tier | Answers follow the permissions already on your files |
| Audit log | Admin actions only, not conversations | Every question, answer and file used |
| Approval before anything leaves | Not found | Every email or post a bot drafts waits for a person |
| Security certifications | Msty says SOC 2 and ISO 27001 are not yet complete | Runs on servers reserved for your firm, see the deployment page |
| Best for | Individuals and small teams who like to tinker | Firms of 20 to 150 with client data to protect |
Fair summary: Msty is the better personal tool. Locul is built for the firm as a whole, which costs more than a desktop licence and gives less room to tinker with every model on Hugging Face.
A desktop AI app only knows the files on one person's laptop. A team workspace knows the shared folders, who may open them, and what happened to each version.
kurze Frage, which version of the Halvorsen framework agreement is the live one?? Lena says v7 but the folder has v7_final, v7_final_FINAL and 'v8 DRAFT do not send'. Customer wants the liability cap clause by 3pm their time, so 2pm ours.
The live one is v7_final_FINAL, signed 2 September 2026. v7_final was replaced the same day after Halvorsen's legal team changed the payment terms in clause 9. 'v8 DRAFT' is a proposal from Lena from October and has not been agreed.
The liability cap is clause 14.2: 150% of the fees paid in the previous 12 months, excluding data protection claims, which are uncapped.
An example. The company, customer and people are invented.
If you are one person, or you have an engineer who wants to run it, one of these may suit you better.
Open-source desktop app for running local models, good for individuals.
Free for work, excellent for trying models on one machine. See our LM Studio comparison.
Strong team features if you can run the server. See our Open WebUI comparison.
Plenty of people will keep a personal AI app on their laptop, and that is fine. What a regulated firm needs on top is one shared place for client work: the same files for everyone, permissions that match the drives, a full log and a person approving anything that leaves. That is the gap Locul fills.
We test open models on real office work: messy emails, spreadsheets with broken formulas, forwarded chains with the ask buried in paragraph four. See how close they get to the big cloud models.
Msty has a free tier and paid tiers, including Enterprise and Teams with a minimum of five seats. Check Msty's pricing page for current terms, as we do not quote other companies' prices.
Yes, on the Enterprise and Teams tier, which adds user and team management, SSO and admin audit logs. Chats stay on each device, and shared resources need S3-compatible storage that you supply.
For one person: Jan or LM Studio. For a technical team that wants to self-host: Open WebUI or AnythingLLM. For a regulated firm that wants one managed workspace on its own or dedicated servers: Locul.
Locul runs leading open-weight models on your private servers and upgrades them as better ones ship. You do not download or manage models yourself.
One call. We set up a demo on files like yours, so you can compare it with Msty side by side.